vvEPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park IMC 27711
EPA-600/9-80-004
January 1980
Proceedings:
Advances in Particle
Sampling and
Measurement
(Daytona Beach, FL,
October 1979)
rivironment
ram Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports m this series result from the
effort funded under the 17-agency Federal Energy/Environment Research and
Development Program. These studies relate to EPA's mission to protect the public
health and welfare from adverse effects of pollutants associated with energy sys-
tems. The goal of the Program is to assure the rapid development of domestic
energy supplies in an environmentally-compatible manner by providing the nec-
essary environmental data and control technology. Investigations include analy-
ses of the transport of energy-related pollutants and their health and ecological
effects; assessments of, and development of, control technologies for energy
systems; and integrated assessments of a wide range of energy-related environ-
mental issues.
EPA REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved
for publication. Approval does not signify that the contents necessarily reflect
the views and policies of the Government, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/9-80-004
January 1980
Proceedings:
Advances in Particle Sampling
and Measurement
(Daytona Beach, FL, October 1979)
W.B. Smith, Editor
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
Contract No. 68-02-3118
Program Element No. EHE624
EPA Project Officer: D. Bruce Harris
Industrial Environmental Research Laboratory
Office of Environmental Engineering and Technology
Research Triangle Park, NC 2771 1
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
PREFACE
The symposium was sponsored by the Process Measurements
Branch, Industrial Environmental Research Laboratory, U.S. En-
vironmental Protection Agency. D. Bruce Harris, EPA, served
as chairman. Session chairmen were Robert Carr (Electric Power
Research Institute), W.B. Kuykendal (EPA), Benjamin Y.H. Liu
(University of Minnesota), Otto Raabe (University of California-
Davis) , and Herbert Spencer, III (Joy Manufacturing Co.). A.B.
Craig, Director of the Industrial Processes Division of EPA,
delivered a welcoming address.
A list of reports on the research supported by the Process
Measurements Branch can be obtained from Mrs. Judy Ford, MD-62,
Process Measurements Branch, Industrial Environmental Research
Laboratory, U.S. Environmental Protection Agency, Research Triangle
Park, NC 27711.
111
-------
CONTENTS
Preface iii
Speakers and Chairmen vii
Paper 1. An Experimental Study of Virtual Impactors
Thomas J. Yule and Christopher G. Broniarek 1
Paper 2. A Theoretical Study of Virtual Impactors
Virgil A. Marple and Chung M. Chien 22
Paper 3. A Heavy Grain-Loading Impactor
Dale A. Lundgren, Ernest R. Cerini and Michael L. Smith 54
Paper 4. Velocimetric Determination of Aerodynamic Diameter
in the Range from 0.1 ym to 15 pm
James C. Wilson and Benjamin Y.H. Liu 67
Paper 5. A Prototype Particulate Stack Sampler with Single-Cut
Nozzle and Microcomputer Calculating/Display System
John C. Elder, Larry G. Littlefield, and Marvin I. Tillery. ... 83
Paper 6. The Effects of Nozzle Losses on Impactor Sampling
Kenneth T. Knapp 101
Paper 7. Dilution Source Sampling System
Robert J. Heinsohn, John W. Davis, and Kenneth T. Knapp 107
Paper 8. Aerosol Characterization with a Quartz Crystal
Microbalance Cascade Impactor
David C. Woods and Raymond L. Chuan 130
Paper 9. Extending Precision in a Computer-Based Cascade
Impactor Data Reduction System
Jean W. Johnson, B. E. Pyle, and Wallace B. Smith 146
Paper 10. Implementing CIDRS - A Programmer's Perspective
Clinton E. Tatsch 167
Paper 11. Numerical Simulation Studies and Data Reduction
for Size Classifying Measurement Techniques
H. Fissan and C. Helsper 180
-------
Paper 12. Non-Ideal Behavior in Cascade Impactors
Joseph D. McCain and James E. McCormack 201
Paper 13. Field Testing of Automatic Piezoelectric Microbalances
for Outdoor Aerosol Mass Concentration Measurements
Kazuo Tsurubayashi and Hajime Kano 217
Paper 14. A New Real-Time Isokinetic Dust Mass Monitoring System
James C.F. Wang 242
Paper 15. A New Real-Time Mass Monitoring Instrument: The TEOM
Harvey Patashnick and Georg Rupprecht 264
Paper 16. New Automated Diffusion Battery/Condensation Nucleus
Counter Submicron Sizing System: Description and Comparison
with an Electrical Aerosol Analyzer
Gilmore J. Sem, Jugal K. Agarwal, and Charles E. McManus 276
Paper 17. An In-Situ Liquid Droplet Sizing System
Daniel E. Magnus and David S. Mahler 302
Paper 18. Some Aerodynamic Methods for Sampling Inhalable
Particles
Wallace B. Smith, Kenneth M. Gushing, M. Christine Thomas,
and Rufus R. Wilson, Jr 316
Paper 19. Deposition of Inhaled Particles and Possible
Sampling Methods
Vittorio Prodi, Giuseppe Tarroni, and Carlo Melandri 348
Paper 20. Aerosol Sampling Inlets and Inhalable Particles
Benjamin Y.H. Liu and David Y.H. Pui 382
VI
-------
SPEAKERS AND CHAIRMEN
Robert Carr
Electric Power Research Institute
P.O. Box 10402
Palo Alto, CA 94304
Phone: 415/855-2422 ext. 137
Raymond L. Chuan
Defense Division
Brunswick Corporation
3333 Harbor Boulevard
Costa Mesa, CA 92626
Phone: 714/546-8030
A.B. Craig
Director, Industrial Processes
Division
Environmental Protection
Agency
Research Triangle Park, NC 27711
Dennis Drehmel
Environmental Protection Agency
Particulate Technology Branch
Utilities & Industrial Power
Division
Research Triangle Park, NC 27711
John C. Elder
Aerosol Studies Section
Industrial Hygiene Group
University of California
Los Alamos Scientific Laboratory
P.O. Box 1663
Los Alamos, NM 87545
H. Fissan
Gesamthochschule Duisburg
Aerosolmesstechnik
4100 Duisburg
Bismarckstrasse, 81
Federal Republic of Germany
Phone: 0203 392 202
D. Bruce Harris
Environmental Protection Agency
Industrial Environmental Re-
search Laboratory
Mail Stop MD-62
Research Triangle Park, NC 27711
Phone: 919/541-2557
Robert J. Heinsohn
Pennsylvania State University
Mechanical Engineering Department
University Park, PA 16802
Phone: 814/865-8281
Jean W. Johnson
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, AL 35205
Phone: 205/323-6592 ext. 547
Kenneth T. Knapp
Particulate Emissions Research
Section
Environmental Sciences Research
Laboratory
Environmental Protection Agency
Environmental Research Center
Mail Stop MD-46
Research Triangle Park, NC 27711
Phone: 919/541-3085
W.B. Kuykendal
Environmental Protection Agency
Industrial Environmental Re-
search Laboratory
Mail Stop MD-62
Research Triangle Park, NC 27711
Phone: 919/541-2557
Benjamin Y.H. Liu
University of Minnesota
Department of Mechanical
Engineering
125 Mechanical Engineering
111 Church Street S.E.
Minneapolis, MN 55455
Phone: 612/373-3043
Dale A. Lundgren
University of Florida
Department of Environmental
Engineering Sciences
A.P. Black Hall
Gainesville, FL 32611
Phone: 904/392-0846
VII
-------
Dan Magnus
KLD Associates, Inc.
300 Broadway
Huntington Station, NY 11746
Phone: 561/549-9803
Virgil A. Marple
University of Minnesota
Mechanical Engineering Department
111 Church Street_S.E.
Minneapolis, MN 55455
Phone: 612/373-9984
Joseph D. McCain
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, AL 35205
Phone: 205/323-6952 ext. 278
Harvey Patashnick
Rupprecht and Patashnick Company
255 Hackett Boulevard
Albany, NY 12208
Phone: 518/438-5569
Vittorio Prodi
Comitato Nazionale per L'Energia
Nucleare
Centro di Calcolo
Laboratorio Fisica Sanitaria
Bologna
Via Massini, 2 - Cap 40138
Italy
Otto Raabe
Laboratory of Radiobiology
University of California - Davis
Davis, CA 95616
Phone: 916/752-7754
Gilmore Sem
TSI Incorporated
500 Cardigan Road
P.O. Box 3394
St. Paul, MN 55165
Phone: 612/483-0900
Wallace B. Smith
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, AL 35205
Phone: 205/323-6592 ext. 520
Herbert Spencer, III
Western Precipitators Division
Joy Manufacturing Company
565 Colorado Boulevard
P.O. Box 2744, Terminal Annex
Los Angeles, CA 90051
Phone: 213/240-2300
C.E. Tatsch
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
Phone: 919/541-6000
Kazuo Tsurubayashi
Nihon Kagaku Kogyo Co., Ltd.
Head Office, 2-1
Shimizu, Suita
Osaka
Japan
James C.F. Wang
Combustion Research Division
Sandia Laboratories
Livermore, CA 94550
James C. Wilson
University of Minnesota
Department of Mechanical
Engineering
125 Mechanical Engineering
111 Church Street S.E.
Minneapolis, MN 55455
Phone: 612/373-9984
David C. Woods
NASA Langley Research Center
Hampton, VA 23665
Thomas J. Yule
Applied Physics Division
Argonne National Laboratory
9700 South Cass Avenue
Argonne, IL 60439
Phone: 312/739-7711
vixi
-------
PAPER 1
AN EXPERIMENTAL STUDY OF VIRTUAL IMPACTORS
THOMAS J. YULE
ARGONNE NATIONAL LABORATORY
AND
CHRISTOPHER G. BRONIAREK
RENSSELAER POLYTECHNIC INSTITUTE
ABSTRACT
Virtual impactors are currently being used in a number of
instruments to separate an aerosol into different size ranges.
The virtual impactor is a variation of the standard impactor
in which the impaction surface is replaced by an orifice into
which particles can pass and be collected or counted. We have
made an experimental study of the collection characteristics
of virtual impactors. The parameters varied included: accelera-
tion nozzle-to-collection probe distance, the ratio of the collec-
tion probe-to-acceleration nozzle diameters, and the ratio of
collection probe-to-inlet flows. Measurements were also made
with different collection probe geometries. It was found that
it is possible to parameterize much of the data by introduction
of the Stokes number and an effective minor flow collection ef-
ficiency. One disadvantage of the virtual impactor is that in
the transition region particles are collected on the inside walls
of the collection probe near the probe tip. The amount that
is collected is a sensitive function of the probe geometry.
INTRODUCTION
Virtual impactors are currently being used in a number of
instruments to separate an aerosol into different size ranges.
The virtual impactor is a variation of the standard impactor
in which the impaction surface is replaced by a collection probe
into which large particles will pass and then be collected or
counted. Figure 1 is a schematic illustrating the operation
of a virtual impactor. The aerosol stream is drawn through the
acceleration nozzle where the velocity of the particles is in-
creased. A collection probe is located at a short distance below
-------
STREAMLINES
TRAJECTORY OF
PARTICLE TOO
SMALL TO IMPACT
ACCELERATION
NOZZLE
/
f
/
/
/
/
/
/
y
W
\ \ i *
QI
(1
1
« DT ••
^ TRA-IFHTORV OP
\
/
r.
f
/
/
/
/
V.
IMPACTED PARTICLE
COLLECTION
-^""^ PROBE
Figure 1. Schematic view of a virtual impactor showing representative streamlines
and particle trajectories.
-------
the acceleration nozzle. Flow conditions are maintained such
that the flow in the collection probe is a fixed fraction of
the inlet flow. A few representative streamlines are shown on
the figure.
In order to understand the principle of operation/ consider
particles that exit from the acceleration nozzle near streamlines
which do not pass through the collection probe. Small particles
follow the streamlines and remain in the flow that does not pass
through the collection probe; large particles are unable to fol-
low the streamlines in regions of rapidly changing curvature
and become caught up in the flow that passes through the collec-
tion probe. The flow in the collection probe, Qj, is referred
to as the minor flow. The ratio Qi/Q<>f where Q0 is the inlet
flow, ranges in value from 0.03 to 0.25 in present designs.
For zero particle size a fraction of the particles equal to Qi/Q0
follow the streamlines of the minor flow and pass through the
collection probe. The small particles that enter the collection
probe along with the large particles introduce cross contami-
nation. The flow which does not enter the collection probe is
referred to as the major flow; it contains most of the small
particles.
Virtual impactors possess a number of advantages over stand-
ard impactors. These are: the size and placement of the size-
separated sample can be optimized for the analysis system, par-
ticle bounce and reentrainment are minimized, and significant
amounts of a sample can be collected without serious loading
problems and without change in collection efficiency. There
are also disadvantages. These are: the slope of the efficiency
curve is less steep than that for a well-designed standard im-
pactor, there is always some cross contamination, and some of
the sample is collected on the inside walls of the collection
probe.
The method of virtual impaction was introduced by Hounam
and Sherwood1 in 1965. They designed and constructed a multi-
stage device called a cascade centripeter, which had cutpoints
of 1.2, 3.5, and 12 micrometers. Flat orifice plates were used
instead of acceleration nozzles. Their device had considerable
wall losses. Shortly afterward, Conner2 investigated the col-
lection efficiency of a single-stage virtual impactor as a func-
tion of operating parameters.
More recently, interest in virtual impactors has been height
tened by the U.S. Environmental Protection Agency's desire for
a device which collects size-segregated ambient aerosol samples
for large-scale monitoring applications. The device, which is
referred to as a dichotomous sampler, separately collects fine
(<2.5 micrometer) and coarse (>2.5 micrometer) airborne particles.
An early design, which has two stages, is described by Dzubay
and Stevens3 and Loo, et al.1* A second generation sampler, which
-------
has only a single stage and very low internal losses, has been
developed by Loo, et al.5 Virtual impactors have also appeared
in other devices. McFarland, et al.6 have developed a sampler
that was used to collect large quantities of particulate matter
from the stack of a coal-fired power plant. Kotrappa, et al.7
and Yule8 have used virtual impactors to size segregate radio-
active aerosols in radiation detection instruments.
To aid in the development of the devices described above,
various experimental studies on virtual impactors were under-
taken. The most extensive studies are those by Loo, et al.4
and by McFarland, et al.9 Both studies were aimed at determining
the optimum parameters for attaining sharp cutoffs with minimum
internal losses for particular systems. The study reported here
is somewhat broader in scope. It was undertaken to provide suf-
ficient information to determine the cutoff size and collection
characteristics for a wide range of system parameters. The study
was patterned after the study on standard impactors by Marple
and Liu.10
There have also been some theoretical studies of virtual
impactors by Marple and Chien11 and by Hassan, et al.12 In gen-
eral, these studies have given insight into the separation charac-
teristics, but the agreement between measured and predicted data
is not as good as that for the standard impactor.
EXPERIMENTAL METHODS
The collection characteristics of virtual impactors were
studied using monodisperse aerosols. Most of the measurements
were made with dioctyl phthalate (DOP) droplets, which contained
trace amounts of uranine (the sodium salt of fluorescein) dye
as a tracer. Aerosols with sizes from 1 to 10 micrometers were
generated with a Berglund-Liu vibrating-orif ice aerosol gener-
ator. Some measurements were also made with polystyrene latex
(PSL) aerosols. The PSL aerosol generator consisted of a nebu-
lizer, diluter, diffusion dryer, and Kr-85 charge neutralizer.
The collection characteristics for the two aerosols are expected
to be different since one is a liquid droplet which will stick
on contacting a surface, while the other is an elastic sphere
which can rebound from a surface. The results obtained with
the DOP aerosols represent an upper limit for collection on
surfaces of the virtual impactor and are easier to compare with
calculated results, because one can assume any droplet that comes
in contact with a surface will stick to that surface.
Figure 2 is a schematic view of the single-stage virtual-
impactor test assembly that was used for the measurements. The
assembly allows one to easily change the acceleration nozzle
and collection probe geometries. For measurements with the DOP
aerosols the quality of the aerosols was monitored with an op-
tical particle counter that had the probe located in a region
-------
till I,<
MAJOR
FLOW
r r { / r r
\
MINOR
FLOW
Figure 2. Apparatus for measuring the collection characteristics of virtual impactors.
Some of the key components are indicated: 1. acceleration nozzle, 2. flow
transition cone, and 3. collection probe.
-------
above the inlet. A Bausch and Lomb 40-lA counter was used with
a modified flow system, as described by Marple, et al.,13 and
with interface electronics that allows pulse-height analysis
with a multichannel analyzer. Such a monitor is very useful
in determining whether the aerosol generator is operating properly.
For the measurements with OOP aerosols filters were inserted
in the minor and major flow lines. Flow measurements were made
with pressure-corrected rotameters. Measurement times were on
the order of 15 minutes. The aerosol depositions were deter-
mined on various comp'onents: the underside of the piece support-
ing the acceleration nozzle, the acceleration nozzle, the collec-
tion probe, and the major and minor flow filters. These are
the only areas where there are significant deposits. The de-
posits were determined by washing the pieces or filters in mea-
sured amounts of a buffer solution (0.05 M Na.HPO^) and using
fluorescence techniques. A Turner Model 110 rluorometer was
used; calibration curves had been determined which relate the
absorption to the uranine concentration. Concentrations as low
as 0.005 yg/ml can be reliably determined.
For measurements with the PSL aerosols two aerosol sizes
were used: 1.10 and 2.02 micrometers. The optical particle coun-
ter probe was placed in a region below the collection probe.
For these measurements only the fraction of the incoming aerosol
that passed through the collection probe was determined. For
a given particle size and flow conditions, the aerosol concentra-
tions before and after each measurement were determined by set-
ting the major flow, Q2, equal to zero and Q0 equal to the nominal
value of Qj. The average concentration, Cg =g, was determined.
With the nominal values of Q0 and Q2 established, the concentra-
tion, C, in the minor flow was measured. The minor flow collec-
tion efficiency, Em, is simply
_
"m (Qo/Qi) CQz=0 '
RESULTS AND DISCUSSION
A series of measurements were made with OOP aerosols in
which various geometrical and flow conditions were varied. A
virtual impactor with geometrical and flow parameters listed
in Table 1 was chosen as a standard; individual parameters were
varied about these values. A value of DJ/DQ = 1.28 was chosen
for the standard, because this value results in minimum wall
losses in the collection probe. The shape of the collection
probe is the same as that shown in Figure 1. As is discussed
below, this shape is not the optimum shape for minimum collec-
tion on the interior walls of the collection probe; however,
it is a shape that has been used in a number of devices and is
easy to treat analytically. Figure 3 shows the collection charac
teristics for the standard. Since the deposits in the accelera-
tion nozzle and on the underside of the piece supporting the
-------
1.2
1.0 —
I I I I
MAJOR FLOW
/"""
I I I I I
Q^QQ = 0.25
Re = 10.000 _
MINOR FLOW
COLLECTION PROBE
I
I
345
PARTICLE DIAMETER,
S/D0 = 1.0
1.28
10
Figure 3. Collection efficiencies for the standard virtual impactor.
-------
TABLE 1. GEOMETRICAL AND FLOW PARAMETERS FOR THE
STANDARD VIRTUAL IMPACTOR
Parameter
Do
D!/DO
S/DO
T/DO
Qo
Qi/Qo
Value
3.912 mm
1.28
1.0
2.0
27.7 S,/mina
0.25
Corresponds to a jet Reynolds number of 10,000.
acceleration nozzle were very small, they are not shown. The
minor flow collection efficiency is 0.25, which is simply the
ratio QJ/QQ, for small particle sizes and reaches a value of
1.00 for large particle sizes. The region in which the minor
flow collection efficiency is varying is referred to as the
transition region. It is useful to define an effective minor
flow collection efficiency, E_,
Em ~ Ql/Qo
Em = 1 - (h/Qo ' (2)
which corresponds to the efficiency for removing particles from
the major flow and having them pass through the collection probe.
Em may assume values from zero to unity. We choose to define
the cutoff for a virtual impactor as that size for which E^ = 0.5,
Note that it is only in the transition region that there is signi-
ficant collection on the walls of the collection probe. For the
standard the maximum value is approximately 0.20. For aerosols
that have a probability of rebounding from contact with a wall,
this maximum is reduced. In order to investigate the wall losses
in the collection probe in more detail, a special collection
probe, shown in Figure 4, was constructed. The thickness of
the annular inserts was one-third D0. Measurements were made
with 2, 2.5, and 3 micrometer DOP aerosols. For all sizes close
to 100% of the deposit appeared on the inside of the top ring.
This result indicates that the shape of the collection probe
could significantly influence the magnitude of the wall losses
in the transition region.
-------
10 mm
Figure 4. Modified collection probe with annular inserts.
For standard impactors it has been found useful to show
the characteristic impaction curves with the abscissa expressed
in units of the square root of the Stokes number, /St, which
is a dimensionless particle size.10 The Stokes number, as de-
fined by Fuchs,11* is the ratio of the particle stopping distance
to the radius of the impactor throat,
St =
pp C Vo Dp
Do/2
(3)
where pp is the particle density, C is the Cunningham slip cor-
rection factor, V0 is the mean velocity in the throat, Dp is
the particle diameter, y is the fluid viscosity, and D0 is the
diameter of the throat. When plotted in this fashion, it has
been found that the impaction curves are only slowly varying
functions of S/D, the Reynolds number, and T/D. (See Figure 1
for definitions of S and T.) A similar representation is useful
for showing the collection curves_for virtual impactors. Fig-
ure 5 shows Em as a function of /St for measurements for which
QI/QO' S/D0, DJ/DO, and T/D0 were held constant and D0, Dp, and
Qo were varied. Two values of D0 were used; 0.3048 and the
standard 0.3912 cm. For a particular data set either Q0 was
-------
OT
COLLECTION EFFICIENCY
p
to
p
b>
p
bo
4.
o o
«;
r* a?
L- ^3>
?Ȥ!
SIS*
I
p
w
p
ui
p
en
p
bo
p
to
-------
varied and Dp kept constant, or Dp was varied and Q0 kept con-
stant. For the former, this corresponds to a variation in Rey-
nolds number, Re, of a factor of four. In general, all the data
for DO = 0.3048 cm fall on a single curve, while those for D0
= 0.3912 cm lie slightly above the curve. This small variation
is almost entirely due to different finishes on the inside walls
of the collection probes. The inside wall of the larger probe
was polished and had a maximum collection of 0.20, whereas the
other one had a rougher surface and had a maximum collection
of 0.25. A typical collection curve for a standard impactor
is also shown on the figure; it is significantly steeper.
The collection characteristics of virtual impactors as a
function of S/D0, D!/DO, and Qi/Qo were determined. The varia-
tion with T/Do was not determined. The analytical studies indi-
cate that it should be small for T/D0 values greater than one-
half. Studies were not made of the variation in collection
characteristics for wide ranges of Reynolds numbers, since the
analytical studies indicate that to see a significant deviation,
it would be necessary to go to very low values of the Reynolds
number. Such measurements would have been quite interesting,
but would have required a complete redesign of the experimental
apparatus. Figure 6 shows the variation in the collection charac-
teristics for 2-micrometer size particles as a function of S/D0,
with all other parameters fixed at those of the standard. This
variation with S/D0 is significantly greater than that seen for
a standard impactor and is related to the divergence of the jet
as S/Do increases. Figure 7 shows the variation in the collec-
tion characteristics for 2 and 10-micrometer size particles as
a function of DJ/DQ. For the 2-micrometer size particles the
collection on the walls of the probe significantly increases
for DI/DO greater than 1.6. Studies by Loo, et al.1* with solid
particles show similar results and also indicate that a mini-
mum occurs at about D!/DO = 1.3. For the 10-micrometer size
particles increased wall losses in the collection probe begin
at DI/DO greater than 1.4.
The ratio Qi/Qo was also varied. Figure 8 shows the results
for Qi/Qo = 0.10. The transition region is moved to larger par-
ticle sizes. The peak of the collection-probe wall-losses curve
stays at around 0.20. There is also a small amount collected
on the underside of the acceleration nozzle and on the accelera-
tion nozzle support structure. Figure 9 shows the results for
Qi/Qo = 0.03. The transition region is moved to even larger
particle sizes. The peak of the collection-probe wall-losses
curve increases to about 0.30. In order to parameterize these
results it is convenient to use the effective minor flow collec-
tion efficiency, Em, which was previously defined. Figure 10
shows Em versus /St for various values of QJ/QQ. It is seen
that over a significant portion of the transition region the
11
-------
1.0
0.8
UJ
£ 0.6
Z
o
O 0.4
8
0.2
COLLECTION
' PROBE
O
S/DO
Figure 6. Variation in the collection efficiencies for 2-micrometer size particles
as a function of S/Drj.
-------
COLLECTION EFFICIENCY
f
N
l.|
ft
.
1
I
•**
I
I
I'
o
o
p
b>
p
bo
fi-
ts)
.
D
D
D
a o
r- o O
m 3, 3,
H TI ~n
HI
•o
3D
O
00
m
-------
frT
COLLECTION EFFICIENCY
f
Po
§
3 *
ft\ ^h
. 03
*
s?
CB
I
•o
>
31
O
|-
m
O
>
m
T=
•>i —
00 —
co —
-------
>
o
ui
O
o
ui
O
o
INLET PLUS
INLET SUPPORT
0.2 -
456
PARTICLE DIAMETER, ju
10
Figure 9. Collection efficiencies for a virtual impactor for which all the parameters
are the same as those for the standard except Q//Q0 = 0.03.
-------
1.0
0.8
0.6
0.4
I 0.2
Q.J/QQ = 0.25
J_
0.03
I
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
\/St~
Figure 10. Effective minor flow collection efficiencies for various values of
data is fitted by a straightJJ.ne. The line may be characterized
by specifying the value of /St at E^ = 0.50 and the rate of
change of the line, R, which is defined as
. 0.80
= 0.20
R =
/St
(4)
E' = 0.50
m
where the subscript on /5t indicates the value of E^. Figure 11
shows the variation of •/St'Em = 0.50 versus Qi/Qo. R remained
constant at a value of 0.65.
In addition to the measurements described above, a number
of measurements were made to determine the effect of changing
the shape of the collection probe. Figure 12 shows two different
shapes of collection probes that were used: a disc with a hole
in it and a tube with rounded inside edges. The collection
characteristics for the disc were identical with those for the
standard. The collection characteristics for the tube with
rounded inside edges were significantly different. The collec-
tion-probe wall-losses curve had approximately the same shape
as that for the standard, but the magnitude was decreased by
about 35%. There was a corresponding change in the collection
curve for the major flow, while the collection curve for the
16
-------
1.25
« 1.00 -
o
ii
E
(.
0.00
0.30
Figure 11. The variation in ^/St at which the effective minor flow collection
efficiency is 0.50 as a function of
E
//
k/vJ
//
(a) DISC WITH A HOLE THROUGH IT
(b) TUBE WITH ROUNDED INSIDE EDGES
Figure 12. Modified collection probes.
17
-------
minor flow remained unchanged. This result indicates that by
rounding the inside edges of the collection probe particles that
would have been deposited near the top are now able to remain
in the major flow. Loo and Cork15 have optimized the geometrical
parameters for a particular design and have been able to reduce
the losses for liquid particles to less than 1% for all particle
sizes up to 20 micrometers.
Measurements were also made of the minor flow collection
efficiency with PSL aerosols. The results are shown in Figure
13. Also shown on the figure are the curves for minor flow and
receiving tube collection with OOP aerosols. Note that the
results are quite similar for values of /S~E less than 0.5, but
that the OOP values are lower than the PSL values over the range
from 0.5 to 0.9. The differences are explained by the fact that
a droplet on striking a wall will stick, whereas a PSL sphere
has a certain probability for rebounding. Apparently, the spheres
that rebound from the walls can either enter the minor flow or
be swept out of the tube in the major flow. The former appears
to be much more likely for spheres with values of /St near 0.8,
the latter for those with values less than 0.5. Loo, et al.1*
found that with solid aerosols of uranine the maximum collection
in the receiving tube with this geometry was only a few percent.
CONCLUSIONS
The purpose of this study was twofold: 1) to provide data
that is useful in designing systems which employ virtual impac-
tors, and 2) to develop an improved understanding of the prin-
ciples of operation of virtual impactors. It was found that
it is possible to parameterize much of the data by the introduc-
tion of the Stokes number and an effective minor flow collection
efficiency. For given Q1/QOI DJ/DQ, and S/D0, there is a single
curve which represents the minor flow collection efficiency as
a function of /St. Analytical studies indicate that the curve
should apply to quite low values of_Reynolds number.11 For dif-
ferent values of Ch/Qo/ Eff, versus /St is rather well represented
by a straight line whose rate of change stays constant and for
which /St Em = o.50 ^s a smoothly decreasing function of Qi/Q0.
Thus, Figures 10 and 11 can be used to determine effective minor
flow collection efficiency curves for a wide range of system
parameters. The study also indicated that the collection ef-
ficiencies for virtual impactors are more dependent on S/D0 for
values greater than one, than the collection efficiency for a
standard impactor.
One disadvantage of the virtual impactor is that in the
transition region particles are collected on the inside walls
of the collection probe near the probe tip. It was found in
this study and in the study by Loo, et al.4 that this collection
is minimized if Dj/D,, is kept at a value of about 1.3. Further-
more, the amount that is collected is a sensitive function of
18
-------
6T
COLLECTION EFFICIENCY
f-o
II Q
-* O'
II
5
II
"I
ii
p
b
p
b
p
to
p
b>
p
oo
p
ro
p
bo
* Od*
10 CO CO bO
o
O
O
3
a
•o
-------
the probe geometry. Rounding the inside of the probe tip, polish-
ing this surface, and maintaining good alignment between the
acceleration nozzle and collection probe minimizes the amount
collected. Using trial-and-error methods, Loo and Cork15 have
been able to minimize the collection peak to less than 1% for
a particular set of flow and geometry conditions. At this time
it is not known whether such dramatic improvements are possible
for arbitrary flow conditions.
ACKNOWLEDGEMENT
This research was performed under the auspices of the U.S.
Department of Energy. The submitted manuscript has been authored
by a contractor of the U.S. Government under contract No. W-31-
109-ENG-38. Accordingly, the U.S. Government retains a non-
exclusive, royalty-free license to publish or reproduce the pub-
lished form of this contribution, or allow others to do so, for
U.S. Government purposes.
REFERENCES
1. Hounam, R.F., and R.J. Sherwood. The Cascade Centripeter:
A Device for Determining the Concentration and Size Distri-
bution of Aerosols. Am. Ind. Hyg. J. 26:122-131, 1965.
2. Conner, W.D. An Inertial-Type Particle Separator for Col-
lecting Large Samples. J. Air Pollut. Control Assoc. 16(1):35-
38, 1966.
3. Dzubay, T.G., and R.K. Stevens. Ambient Air Analysis with
Dichotomous Sampler and X-Ray Fluorescence Spectrometer.
Environ. Sci. Technol. 9(7):663-668, 1975.
4. Loo, B.W., J.M. Jaklevic, and F.S. Goulding. Dichotomous
Virtual Impactors for Large-Scale Monitoring of Airborne
Particulate Matter. In: Fine Particles: Aerosol Generation,
Measurement, Sampling, and Analysis, B.Y.H. Liu, ed. Academic
Press, New York, 1976. pp. 311-350.
5. Loo, B.W., R.S. Adachi, C.P. Cork, F.S. Goulding, J.M. Jaklevic,
D.A. Landis, and W.L. Searles. A Second Generation Dichotomous
Sampler for Large-Scale Monitoring of Airborne Particulate
Matter. Lawrence Berkeley Laboratory Report LBL-8725, 1979.
6. McFarland, A.R., R.W. Bertch, G.L. Fisher, and B.A. Prentice.
Fractionator for Size Classification of Aerosolized Solid
Particulate Matter. Environ. Sci. Technol. 11(8):781-784,
1977.
7. Kotrappa, P., D.P. Bhanti, S.K. Dua, and P.P. Joshi. A
Single Stage Centripeter for Rapid Analysis of Long-Lived
Alpha Emitters in Air. Health Phys. 27:103-108, 1974.
20
-------
8. Yule, T.J. An On-Line Monitor for Alpha-Emitting Aerosols.
IEEE Trans. Nucl. Sci. NS-25(1):762-766, 1978.
9. McFarland, A.R., C.A. Ortiz, and R.W. Bertch, Jr. Particle
Collection Characteristics of a Single-stage Dichotomous
Sampler. Environ. Sci. Technol. 12(6):679-682, 1978.
10. Marple, V.A., and B.Y.H. Liu. Characteristics of Laminar
Jet Impactors. Environ. Sci. Technol. 8 (7):648-654, 1974.
11. Marple, V.A., and C.M. Chien. A Theoretical Study of Virtual
Impactors. University of Minnesota, Particle Technology
Laboratory Publication 378, 1978. Submitted to Environ.
Sci. Technol.
12. Hassan, Y.A., B.C. Jones, and T.J. Yule. An Analytical
Study of Virtual Impactor Aerosol Separators. To be pub-
lished in Trans. Am. Nucl. Soc., San Francisco Meeting,
1979.
13. Marple, V.A., N.J. Barsic, and K.T. Whitby. Instruments
and Techniques for Dynamic Particle Size Measurements of
Coal Dust. University of Minnesota, Particle Technology
Laboratory Publication 215, 1974.
14. Fuchs, N.A. The Mechanics of Aerosols. Pergamon Press,
New York, 1964.
15. Loo, B.W., and C.P. Cork. High-Efficiency Virtual Impactors.
Submitted to Environ. Sci. Technol.
21
-------
PAPER 2
A THEORETICAL STUDY OF VIRTUAL IMPACTORS
VIRGIL A. MARPLE
CHUNG M. CHIEN
PARTICLE TECHNOLOGY LABORATORY
MECHANICAL ENGINEERING DEPARTMENT
UNIVERSITY OF MINNESOTA
ABSTRACT
The characteristics of virtual impactors have been deter-
mined by the numerical solution of the Navier-Stokes equations
and of the equations of motion of the particles. The effect
of the nozzle Reynolds number, the fraction of flow passing
through the collection probe, collection probe diameter, nozzle
throat length, nozzle-to-collection probe distance, and collec-
tion probe inlet design on the small and large particle collec-
tion efficiencies has been studied. In addition, it was found
that some particles would impact on the inner surface of the
collection probe. The results show that most parameters, with
the exception of the nozzle Reynolds number, have little effect
on the large particle collection efficiency. However, the effect
on the small particle collection efficiency and collection probe
losses was significant for many of these parameters.
INTRODUCTION
The virtual impactor is a device used for the inertial sepa-
ration of airborne particles.1'2 In this impactor, shown sche-
matically in Figure la, a jet of particle-laden air is directed
at a collection probe which is slightly larger in diameter than
the acceleration nozzle. The large particles cross the air
streamlines and enter the collection probe, while the small par-
ticles follow the air streamlines into the side passage. To
remove the large particles from the collection probe, a fraction
of the total flow passing through the nozzle of the virtual
impactor is allowed to pass through the collection probe. This
flow will be referred to as the minor flow, while the flow through
the side passage will be referred to as the major flow.
22
-------
TOTAL FLOW
MAJOR
FLOW
TRAJECTORY OF
PARTICLE TOO
SMALL TO BE
COLLECTED
TRAJECTORY OF
COLLECTED
PARTICLE
COLLECTION PROBE
MINOR
FLOW
(a) VIRTUAL IMPACTOR
100
u
REGION 1
100% -SMALL
PARTICLE
COLLECTION
EFFICIENCY
REGION 3
LARGE PARTICLE
COLLECTION
EFFICIENCY
IDEAL LARGE
"PARTICLE
COLLECTION
EFFICIENCY
PARTICLE DIAMETER, Dp
REGION 1 - PARTICLES IN THE MAJOR FLOW
REGION 2 - PARTICLES IMPACTED ON COLLECTION PROBE
REGION 3 - PARTICLES IN THE MINOR FLOW
(b) COLLECTION EFFICIENCY CURVES
Figure 1. Nomenclature, streamlines, particle trajectories, and efficiency
curves for a virtual impactor.
23
-------
As is the case with real impactors, the virtual impactor's
performance is characterized by a collection efficiency curve.
For the ideal virtual impactor, the separation between the large
and small particles should be perfectly sharp, as shown for the
ideal case in Figure Ib. Note, however, that in the virtual
impactor, there will always be some of the small particles in
the large particle fraction due to the air flow through the
collection probe.
In an actual virtual impactor, the efficiency curve is not
quite so simple, since there are not only large particles passing
through the collection probe and small particles passing through
the side passage, but also a fraction of the particles impacting
upon the inner surfaces of the collection probe. Thus, as shown
in Figure Ib, there are actually two efficiency curves separating
the particles into the following three regions: (1) small par-
ticles passing out the side passage, (2) particles which are
impacted upon the collection probe (losses), and (3) large par-
ticles passing through the collection probe. Since losses in
the virtual impactor are highly undesirable, an impactor should
be designed such that the displacement of the two efficiency
curves shown in Figure Ib is as small as possible, with the two
efficiency curves coinciding for the desirable case of a virtual
impactor with no losses.
It is the purpose of this paper to use numerical methods
to determine the flow fields, particle trajectories, and finally,
the efficiency and loss curves for virtual impactors operating
at different conditions of Reynolds numbers, fraction of flow
passing through the collection probe, and virtual impactor de-
sign.
THEORETICAL TECHNIQUES
The method used to theoretically analyze the performance
of the virtual impactor is to first determine the flow field
within the impactor by solving the Navier-Stokes equations using
numerical analysis techniques, and then to solve for the particle
trajectory within this flow field by numerically integrating
the particle's equation of motion. This method has been used
successfully in a fundamental study of real impactors to deter-
mine the influence of various parameters on the characteristic
collection efficiency curves.3 Comparisons of the theoretical
efficiency curves with those of experimental investigationslt~8
have shown that the agreement is good if the impactor inlet con-
ditions, shapes, and Reynolds numbers are similar. Since the
flow field and particle trajectories are similar in real and
virtual impactors, this theoretical technique should be equally
successful in determining the efficiency curves of virtual im-
pactors.
24
-------
The general method of solution of the flow field is to first
express the Navier-Stokes equations in terms of the vorticity
and the stream function. The resulting differential equations
are then made dimensionless, with the radial, r, and axial, z,
dimensions being in units of the nozzle throat diameter, Dfl.
The Reynolds number ,
Re = P D° V° (1)
y
where p is the fluid density, VQ is the mean fluid velocity at
the nozzle throat, and y is the absolute viscosity of the fluid,
will be a parameter in these equations. The dimensionless dif-
ferential equations are next expressed in a finite difference
form and solved by the method of relaxation over a grid of node
points covering the field of interest, as shown in Figure 2a.
(Note that the flow is symmetrical about the centerline.) Although
the solution is the determination of the vorticity and stream
function values at each node point, the values of the velocity
components at the node points can be calculated from the stream
function values. For details on the derivation of the finite
difference equations, boundary conditions, and relaxation tech-
nique, the reader is referred to a previous paper by Marple et al.9
After the flow field has been determined, it is then neces-
sary to follow particle trajectories through the virtual impac-
tor. To accomplish this, the particle's equations of motion
in the r and z directions are made dimensionless by again express-
ing the r and z dimensions in units of D0.* For these equations,
the Stokes number, St, defined by Fuchs1Q as the ratio of the
particle stopping distance to D0/2, will be a parameter. The
Stokes number is thus expressed as
P V0 C D 2
St = P P (2)
St 9 M D0 U}
where pp is the particle density, C is the Cunningham slip cor-
rection, and Dp is the particle diameter. Since St is dimension-
less, Equation 2 indicates that /St is a measure of the dimen-
sionless particle diameter. The dimensionless equations of
motion are next put in finite difference form and integrated
numerically through the area of interest. This technique, which
is described in detail by Marple and Liu,1* is capable of describ-
ing the particle's trajectory once the particle's initial posi-
tion and velocity have been given.
The integration_process is started by first assigning a
specific value of /St to a particle, and giving the particle
an initial velocity equal to the local fluid velocity at a position
25
-------
near the entrance. By use of the Runge-Kutta integration method,
the movement of the particle, Az and Ar , during a small increment
of time, At, is determined. This gives the position of the par-
ticle at the end of the time increment. This process is then
repeated, and the movement of the particle through the impactor
is followed until the particle either exits through the collec-
tion probe, exits to the side of the virtual impactor, or impacts
upon the wall of the collection probe.
The particle is considered impacted when the center of the
particle comes within one particle radius, rp, of a surface.
Since all dimensions of the virtual impactor are in units of
the nozzle diameter, D0, the particle radius must also be expressed
in units of D0. Thus, from Equations 1 and 2,
JB. - J4-5 P st
Do f Pp Re
A detailed description of the numerical procedure and the com-
puter program used has been given by Marple.3
RESULTS
As described above, the flow fields, particle trajectories,
and corresponding efficiency curves of a virtual impactor will
depend on the parameters Re and Q!/QO and the physical design.
Referring to Figure 1, the design can be specified by the values
of DI/DO, LQ/DO, S/D0f and 00, and the shape of the entrance
to the collection probe.
To initiate the parametric study, a set of base values for
the parameters were chosen and are listed in the "boxed in" por-
tion of Table 1. The other values of the parameters in Table 1
were then varied one at a time to the values listed, while the
remaining parameters remained at the base values. Following
is a discussion of the results for the base case, and then dis-
cussions of the effects of the various parameters.
Base Case
The grid used for the base case, corresponding to the param-
eters in Table 1, is shown in Figure 2a. A thin-walled tube
collection probe, which simulates a collection probe with its
wall tapered to a sharp edge at the entrance, was chosen, since
this design introduces no new variables such as wall thickness,
or radius of curvature of the probe entrance.
The flow field streamlines for the base case are shown in
Figure 2b for streamlines corresponding to 5%, 10%, 20%, 40%,
80%, and 100% of the flow inside that streamline. Note that
26
-------
the 0% streamline is at the centerline, the 10% streamline inter-
sects the collection probe (Qi/Q0 = 10%), and the 100% stream-
line corresponds to the nozzle wall. Also note that the free
streamline emitted from the nozzle reattaches to the surface
defining the nozzle exit plane, forming a recirculation between
that surface and the free streamline. A second recirculation
region is formed between the air flowing out the side passage
and the lower surface, causing a 5% streamline to be indicated
in this region.
In Figure 3a, the particle trajectories for five particles
with different values of /St, all starting at the 50% streamline,
are shown. These five particle trajectories include the three
cases where particles (1) pass through the collection probe,
(2) are impacted on the collection probe inner surface, and (3)
pass through the side passage. Also included are the two critical
trajectories^ between these three cases. For example, if the
value of /ST of a particle is greater than 0.881, the particle
will pass through the collection probe; if it is less than 0.677,
the particle will pass through the side exit; and if it is be-
tween these two critical values, the particle will impact on
the collection probe inner surface.
TABLE 1. VALUES OF DESIGN PARAMETERS USED IN STUDY
Re
Qi/Q.
Lo/D,
S/D,
e,
Collection
probe3
1
10
100
(B) =| |
(C) — ' !
1,
5,
15,
500
000
000
000
0.
0.
0.
0.
05
10
15
25
1.
1.
1.
16
33
49
1 1
0.013 0.25 30° (D) ^ j
2.5 1 45° (A)~~||
2
b
Collection probe A - thin wall
Collection probe B - infinite wall thickness
Collection probe C - finite wall thickness
Collection probe D - finite wall thickness with taper
(Collection probe designs are shown in Figure 18.)
Base values from which parameters are varied.
27
-------
00
5%
10%
20%
40%
80%
100%
RECIRCULATION
REGION-^
• . i. i
RECIRCULATION
REGION
NOZZLE
EXIT PLANE
(a) GRID
(b) STREAMLINES FOR BASE CASE
Figure 2. Grid and flow field for virtual impactor at base case conditions.
-------
to
VD
PARTICLES STARTING
AT 50% STREAMLINE —J
J .,
y/lt
0.575
0.677
0.779
0.881
0.983
PARTICLE
TERMINATION
MAJOR FLOW
CRITICAL
IMPACT ON PROBE
CRITICAL
MINOR FLOW
PARTICLES STARTING
STREAMLINE
(a) FIVE PARTICLES OF DIFFERENT SIZES
STARTING AT THE SAME POINT
(b) CRITICAL TRAJECTORIES FOR
PARTICLES OF SIZE V^St"= 0.5
Figure 3. Particle trajectories at base conditions.
-------
Another method by which the critical values of /St can
be determined is to keep /St constant while varying the inlet
starting position of the particle, as shown in Figure 3b. This
shows that particles with /St = 0.5 starting between the 24%
streamline and the centerline will pass through the collection
probe, those starting between the 24% and 75% streamlines will
impact on the collection probe, and those starting at streamlines
greater than 75% will pass out the side exit.
By using data such as represented in Figure 3a for particles
starting at several positions, or as in Figure 3b for particles
with different values of /St, the "large particle collection
efficiency" and the "small particle collection efficiency" curves
shown in Figure 4 can be determined. In either case, the col-
lection efficiencies are the percent of particles issuing from
the nozzle which pass through the collection probe or side pas-
sage, respectively. The value of /St as a function of these
efficiencies is presented as the base case in the two tables
of the Appendix. Also note that this information for all cases
listed in Table 1 is presented in the Appendix.
For the purpose of determining the percent of the particles
being impacted on the inner surface of the collection probe,
referred to as "collection probe loss" in this paper, it is best
to construct the curve "100% - small particle collection effici-
ency". This curve, along with the large particle collection
efficiency curve, represents the two efficiency curves in Figure Ib.
As seated before, the difference in the efficiencies at any value
of /St represents the collection probe loss, and thus the loss
curve can be constructed.
It should be noted in Figure 4 that the collection charac-
teristics of a virtual impactor can be specified by the large
particle collection efficiency curve and any one of the other
three curves. However, since the collection probe losses are
of considerable importance, this curve will be used in this paper.
Also note that the quantity of collection probe losses is quite
large, being greater than has been reported in experimental in-
vestigations.11'12 However, investigations of the computer out-
puts indicated that nearly all losses occurred at the tip of
the collection probe by particles that were traveling vertically
upward directly adjacent to the probe inner surface. As shown
later, by proper contouring of the probe entrance, it is possible
to reduce these losses. Therefore, the absolute values of the
losses in this paper may be high, but the relative effects of
the various parameters on these losses are of interest.
Although the contour of the collection probe entrance af-
fects the probe losses, it has little effect on the large par-
ticle collection efficiency curve. Thus, this curve should be
considered as the most significant characteristic curve for
virtual impactors, and it is the curve by which virtual impactors
will be characterized in this paper.
30
-------
100
80
CO
CO
O
O
Z
<
O
ill
O
O
Ul
O
O
60
40
20
T
100% -SMALL PARTICLE
COLLECTION EFFICIENCY
CD-
SMALL PARTICLE
COLLECTION EFFICIENCY
Re = 5000
0-,/Qo = 0.10
1.33
2.5
s/D0 = 1
00 = 45°
0
0.1
LARGE PARTICLE
COLLECTION
EFFICIENCY
FRACTION IMPACTED
ON COLLECTION
PROBE (LOSS)
I I I I
0.5
1.0
2.0
Figure 4. Collection efficiency and loss curves for base conditions.
It is of interest to compare the large particle collection
efficiency curve determined theoretically to experimental curves
by other investigators in Figure 5. The values of the parameters
listed in Figure 5 for each curve indicate that the virtual im-
pactors are similar. The comparison shows that the theory agrees
well with the experimental curves of Loo11 and McFarland et al.12
The theoretical curve is nearly identical to Loo's experimental
curve, with the experimental curve of_ McFarland et al. having
approximately 20% lower values of /St.
Influence of Reynolds Number
To determine the influence of the Reynolds number, cases
were run for the different values of Re listed in Table 1, while
the other parameters were held constant at base values. The
flow fields are shown in Figure 6 for Re = 1, 10, 100, 500, 1,000,
and 15,000, and in Figure 2b for Re = 5,000. In Figure 6, only
the portions of the flow fields in the vicinity of the collection
probe are shown, since the flow fields in other portions of the
impactors are very similar to those in Figure 2b. It should
be noted that the flow may not be laminar for Re = 15,000, but
is presented as a limiting case.
31
-------
100
80
n- 60
o
LLJ
o
LL
LL
UJ
20
THEORY
PRESENT WORK
EXPERIMENTAL
LOQ11
McFARLANDl2
0.1
0.5
1.0
2.0
\/~St~
Figure 5. Comparison of theoretical and experimental large particle collection
efficiency curves at the following conditions:
Re QJ/QO s/D0 DJ/D
Present theory 5000 0.1 1.0 1.33
Looll 6000 0.1 0.8 1.38
McFarland12 4500 0.1 0.9 1.31
2.5
0.8
e0
450
450
Since the large particle collection efficiency curves are
of primary interest, the flow fields within the collection probe
are of most importance. It is in this region where we observe
the unexpected result that the penetration of the 10% and 20%
streamlines into the collection probe is greater for the cases
of Re = 100 and 500 instead of for the higher values of Re, where
the inertial effects of the flow should be larger. The reason
for this can be seen by inspecting the velocity profiles at the
nozzle exit plane shown in Figure 7. At high values of Re (Re =
5,000 and 15,000), the velocity profile is relatively uniform
across much of the nozzle, making it difficult for any portion
of the flow to penetrate into the air in the collection probe.
For the cases of Re = 500 and 100, the velocity profile is more
parabolic, and the relatively high velocity near the centerline
makes it possible for the jet to penetrate farther into the col-
lection probe. However, if the value of Re is small (Re = 10),
there is insufficient inertia in the jet to penetrate into the
32
-------
:si
sli
••*
.-••*
Re = 1
U)
Re = 500
•/.'S
» it*
I*
° i»
lit
iVw^
VJ!
ii
..•••::"j»gs?
,' ** « -V
Re= 10
• l»
* A i * ***ii>ii%
^H
-.•^
Re = 1000
4 •**!
* "•»
» •**
4 •**
y ::s
• • •*«
'/i--
ufe
U?
; f
* »
0
.
Re = 100
Re = 15,000
a ,-aj
/
:L
*i /I-*-
- ,v|
/. I''
Figure 6. Theoretical streamlines at the specified values of Re
= o. 1, DJ/DQ = 1-33, LO/DO = 2.5, S/DO = 1, and e0 = 45°).
-------
DISTANCE FROM AXIS, r
0.5 0.4 0.3 0.2 0.1
Figure 7. Influence of the Reynolds number on the axial velocity profile at
the nozzle exit.
34
-------
viscous air in the collection probe. Thus, Re in the range of
100 to 500 is the correct combination of nozzle velocity profile
and inertia to obtain maximum penetration of the 10% and 20%
streamline into the collection probe.
The large particle collection efficiency curves and loss
curves are shown in Figure 8 with Re as a parameter. The in-
fluence of the flow penetration into the collection probe can
be seen in the position of the large particle collection effi-
ciency curves in Figure 8a. For the cases where the flow pene-
tration into the collection probe is largest (Re = 100 and 500),
the particle Stokes numbers must be larger in order for the par-
ticles to penetrate into the minor flow, since the Stokes number
is defined as the ratio of the particle stopping distance to the
radius of the nozzle. Where the flow penetration into the collec-
tion probe is small, the required distance that the particle must
travel, and thus the value of the Stokes number, will be smaller.
It is of interest to note in Figure 8a that the efficiency
curves are nearly identical for the cases of Re = 1 and 10, and
for Re = 5,000 and 15,000, indicating that lower or higher Rey-
nolds numbers than those listed should have little effect on these
curves. It is also interesting to note that the slopes of the
penetration curves are greater for the larger values of Re, in-
dicating better cut-off characteristics. However, this effect
is small, which is different from real impactors, where the cut-
off characteristics are much poorer for low values of Re than
for large values.1*
Concerning the influence of Re on the losses, Figure 8b
shows that the influence is large. In general, the losses in-
crease from a minimum of about 10% at low values of Re to about
60% at high valjjes of Re, with the maximum losses occurring at
the value of /St corresponding to 50% efficiency.
Influence of Qi/Q0
The flow fields corresponding to Qj/Qo values of 0.05, 0.15,
and 0.25 are shown in Figure 9, and should be compared to the
base case of Qi/Q0 = 0.10 in Figure 2b. As can be seen from
these figures, more streamlines pass through the collection probe
for large values of Ch/Qo and the small amount of flow in the
side passage leaves more room for recirculation in this area.
Tne flow field for Q^/Q0 = 0.05 shows reattachment to the lower
surface of the side passage, much like the flow at Re = 100 in
Figure 6. Whether or not the flow field in this region is cor-
rect is uncertain. However, since the flow in the region has
no effect on the flow field in the important area within the
collection probe, a more detailed investigation of the flow in
this region was not made.
The corresponding large particle collection efficiency curves
and loss curves are shown in Figure 10. The large particle col-
lection efficiency curves indicate that a "sharper" cut between
35
-------
oo
I
Po
S £
o
o
I-
I-
a $
I l
o
00
m
II Q.
\o
O=s
5
Co
r-
O
vt
v>
p
is)
COLLECTION PROBE LOSS, %
8 § g §
fO
o
8
-------
w
p
o
o
ii
p
N9
a\
p
en
""""°"°^aSSaSiSSiHiiJHjS 8_8_8
pew **"«* •»•*««••»*• •• •« • * •* »•
•*"*aai?J8S«83 j i S si H 1 1 88
O
o
ii
p
b
(D
or sc oc
Figure 9. Theoretical streamlines at the specified values of Q 1/Q.Q (Re = 5000,
= 1.33, LQ/DO = 2.5, S/D0 = 1, and 60 = 45o).
-------
I
CD
^ ff
- S e* —
»a.8. g
r~
35 «•
§
'
03?
—» O
Q)
8
*>§'
-* --3
b|
^ I
gf
r-S
%i
CJ
p
ro
a
o
-o
yj
O
m
m
O <$
p
VI
COLLECTION PROBE LOSS, %
3 B
— 1 \
8
T~
o
o
p p p p
O W O
Pi
-------
large particles collected and those which are not is obtained
for smaller values of Qi/Qo- Also, the cut-off size increases
as Qi/Qo decreases. This is expected, since particles must pass
through more air to enter the minor flow stream when Qi/Q0 is
small.
The loss curves show more losses associated with the lower
Qi/Qo values. This would be due to the larger percentage of
the flow being exposed to the inlet of the collection probe where
the losses generally occur, and should be decreased with proper
inlet design.
Effect of Pi/Dp
Besides DJ/DQ being equal to the base value of 1.33 (Fig-
ure 2b), D!/DO was also set at 1.16 and 1.49 (Figure 11). Al-
though, as shown in Figure 12, this parameter had only a small
effect on the large particle collection efficiency and loss
curves, there are substantial differences in the flow fields
for these three cases. For example, the flow field attaches
close to the nozzle exit when D!/DO = 1.49 but does not attach
when D!/DO = 1.16. Experiments (personal communication from
B.W. Loo, Lawrence Berkeley Laboratories) have shown that for
values of D!/DO on the order of 1.49, the reattaching flow does
cause particles to impact on the nozzle exit plane surface,
increasing the losses in the virtual impactor. Therefore, it
is recommended that the value of Di/D0 be kept less than 1.49
and preferably near 1.33.
Effect of L0/D0
To investigate the effect of L0/D0f the results of a case
where L0/D0 is small (L0/Do = 0.013) is compared to the base
case. Although the streamlines issuing from the nozzle are
not parallel to the nozzle axis when L0/D0 = 0.013 (Figure 13)
as they were when L0/D0 =2.5 (Figure 2b) , the effects on the
large particle collection efficiency and the loss curves are
negligible, as shown in Figure 14. Similar insensitivity to
this parameter was found for real impactors.
Influence of
The influence of S/Do was determined by using values of
S/DO = 0.25, 1 (base), and 2. The flow fields shown in Figures
15 and 2b appear to be quite different. However, in the region
in the collection probe where the 10% streamline attaches to
the probe wall, the flow fields are similar. Thus, as expected,
the resulting large particle collection efficiency curves shown
in Figure 16 are nearly identical. Again, this is different
from the effect found for real impactors ** where small values
of S/D0 have a large effect on collection efficiency.
39
-------
i i 4
1.16
A «*» I
A ««• L
A ««V j»
•«w|
JA «*W
JA ••«•
/* •*»
|y.*"
^f*
:
-------
COLLECTION PROBE LOSS,
So
;n e
§.§
Q.
8
§
01
a
o
o
00
m
V)
p
U1
ro
b
g
§
8
a
o
o
o
m
•o
o
r-
rn
o
O
r-
r-
m
o
O
z
m
m
O
EFFICIENCY, E (%)
S §
*» w -»
(O W O)
t> O D
a
o
-------
15 10 .5 1
/5. Theoretical streamlines at LQ/DQ= 0.013 (Re = 5000, QJ/QQ = 0.1,
= 1.33, S/D0 = 1, and 60 = 45°).
42
-------
100
55
lif
o
LU
O
EZ
LL
LU
20 -
V)
8
LU
03
O
cc
a.
O
o
0.5 1.0
V^st"
(a) LARGE PARTICLE COLLECTION EFFICIENCY
100 i 1 1 1 1 1—I—n
80
60
40
20
2.0
0.2 0.5 1.0
\/St~
(b) COLLECTION PROBE LOSS
2.0
Figure 14. Large particle collection efficiency and collection probe loss curves
at the specified values of LQ/DQ (fie = 5000, Qj/Qn = 0.1, DJ/DQ = 1.33,
S/D0= 1, and Q0
43
-------
» 5 s s ; ;.;*•:>.
= 0.25
i i •
= 2
Figure 15. Theoretical streamlines at the specified values of S/DQ (Re = 5000,
= 1.33, L/D = 2.5, and Q= 45O).
= 0. 1,
44
-------
100
80
60
u
>
u
40
LL
U.
HI
20
S/DQ
0.25 O
1 O
2 A
0.2
0.5
1.0
2.0
(a) LARGE PARTICLE COLLECTION EFFICIENCY
100
80
CO
8
ui
§ 60
cc
a.
O
O
40
20
i I *
I
S/Dr
0.2
0.5
1.0
2.0
(b) COLLECTION PROBE LOSS
Figure 16. Large particle collection efficiency and collection probe loss curves at the
specified values of S/Drj (Re = 5000, G//Q0 = °-1> Dj/D0 = 1-33, LQ/DQ = 2.5,
and do = 450).
45
-------
However, the probe loss curves shown in Figure 16 are in-
fluenced by S/Do, with larger losses being found with small S/D0
values. It may be expected that additional losses will be ex-
perienced on the nozzle exit plane surface and on the backside
of the collection probe for small S/D0 values due to turbulence
in the restricted area between the collection probe and nozzle
exit plane.
Influence of 00
The effects of the entrance angle on the large particle
collection efficiency and probe loss curves are shown in Fig-
ure 17. Since the streamlines are very similar to the base case
shown in Figure 2bf the streamlines for 60 = 30° are not shown.
From Figure 17, it can be seen that the large particle col-
lection efficiency curves are similar in shape for both angles,
but the curve for 00 = 45° is shifted to smaller particle sizes.
This is due to the particles being thrown closer to the center-
line for the larger 90 values, making the collection of particles
in the probe slightly easier. The losses shown in Figure 17
indicate less loss for 00 = 45° than for 60 = 30°. Thus, it
appears that 00 should be at least 45°.
Influence of the Collection Probe Inlet Design
Four receiving tube configurations were tested as shown
in Figures 2b and 18. The base design was a thin wall shown
as configuration A. In configuration B, the wall was infinite
in width, and in configuration C, the wall had a finite thickness,
In configuration D, the wall thickness was also finite, but the
inner surface was tapered in an attempt to reduce losses.
The flow fields for these configurations are shown in Fig-
ure 18, and the resulting large particle collection efficiency
and probe loss curves are shown in Figure 19. Since configura-
tion B is essentially a real impactor impaction plate with a
hole in its center, particles are collected on the plate, and
thus losses are meaningless for this case and are not shown.
It is interesting to note that the configuration of the
entrance to the collection probe had essentially no effect on
the large particle collection efficiency curve. This would be
expected, since this curve defines the separation of particles
which impact upon the probe wall and those which pass .through
with the minor flow, and since the region of importance for this
curve is inside the collection probe rather than at its entrance.
The losses, however, are influenced by the collection probe
configuration. For configurations where there is a sharp edge
at the upper entrance, such as configurations A and C, the probe
losses are quite large and the loss curves are similar.
46
-------
100
80
GO
u
—
o
4- 40
20
1 r
30°
45°
0.2
0.5
1.0
2.0
-------
I 1 1 1 1 . I 1 . . 1 . 1 H fi i ±1 F//5
COLLECTION PROBE B
1 ' i » I
COLLECTION PROBE C
-------
I
5;
O -Q JB
:i 3 §.
-»s-9
.-^
&
to
NJ
o
o
o
CO
o
o
o
m
§ <
? »i
p
Ol
v>
V)
N)
b
D
-L
>
-L
o
• L.
m
n
O
00
m
o
o
p
ro
33
O
m
Z
73
H
a
t-
m
O
m
00
•L - L JL -L
a
o
r-
i-
m
n
-------
The losses that were found in these cases are from particles
which are traveling vertically upward along the collection probe
wall toward the major flow exit. The boundary layer in this
region is very thin, and particles passing within one particle
radius of this wall are collected as a loss. If there were some
mechanism, such as aerodynamic forces, which would keep the par-
ticle from touching the wall, it would not be collected as a
lost particle.
In addition, losses should be reduced by replacing the sharp
corner in configuration C with a tapered entrance (configuration
D) so that the particles could not be collected as easily at
the upper corner of the probe. As shown in Figure 19b, the
losses were reduced. By replacing the taper with a radius, the
losses should be reduced even farther.
CONCLUSIONS
An examination of the large particle collection efficiency
curves reveals that they all have essentially the same shape
and are not greatly influenced by any of the parameters. The
only parameter which appears to influence the shape is Ql/Q0,
because the collection efficiency curves are_ajsymptotic to the
different values of Qi/Qo at low values of /St.
The reason none of the parameters has a large effect on
the large particle collection efficiency curve is that this ef-
ficiency is governed by the flow field within the collection
probe. The losses, however, are governed by the flow conditions
at the tip of the collection probe inlet, and thus are influenced
by many of the parameters. This is especially true for the case
of the collection probe inlet design, where losses were reduced
by adding a taper to the probe inlet.
ACKNOWLEDGEMENT
This work was supported under Bureau of Mines Contract H0177026
through the Twin Cities Mining Research Center. The financial
support of the Bureau is gratefully acknowledged. This report
is Particle Technology Laboratory Publication No. 378.
REFERENCES
1. Loo, B.W., J.M. Jaklevic, and F.S. Goulding. Dichotomous
Virtual Impactors for Large-Scale Monitoring of Airborne
Particulate Matter. In: Fine Particles: Aerosol Genera-
tion, Measurement, Sampling, and Analysis, B.Y.H. Liu, ed.
Academic Press, New York, 1976. pp. 311-350.
2. Dzubay, T.G., and R.K. Stevens. Ambient Air Analysis with
Dichotomous Sampler and X-Ray Fluorescence Spectrometer.
Environ. Sci. Technol. 9:663-8, 1975.
50
-------
3. Marple, V.A. A Fundamental Study of Inertial Impactors.
Ph.D. Thesis, University of Minnesota, Mechanical Engineer-
ing Department, Particle Technology Laboratory Publication
144, 1970.
4. Marple, V.A., and B.Y.H. Liu. Characteristics of Laminar
Jet Impactors. Environ. Sci. Technol. 8:648-654, 1974.
5. Jaenicke, R., and I.H. Blifford. The Influence of Aerosol
Characteristics on the Calibration of Impactors. J. Aerosol
Sci. 5:457-464, 1974.
6. Willeke, K., and J.J. McFeters. The Influence of Flow Entry
and Collecting Surface on the Impaction Efficiency of Inertial
Impactors. J. Colloid Interface Sci. 53:121-7, 1975.
7. Schott, J.H. Jet-Cone Impactors as Aerosol Particle Sepa-
rators. M.S. Thesis, University of Minnesota, 1973.
8. Willeke, K. Performance of the Slotted Impactor. Am. Ind.
Hyg. Assoc. J. 36:683-691, 1975.
9. Marple, V.A., B.Y.H. Liu, and K.T. Whitby. On the Flow
Fields of Inertial Impactors. J. Fluids Eng. 96:394-400,
1974.
10. Fuchs, N.A. The Mechanics of Aerosols, Pergamon Press,
New York, 1964. p. 154.
11. Loo, B.W., and C.C. Cork. High Efficiency Virtual Impactor
for Collecting Airborne Particulate Matter. Lawrence Berkeley
Laboratory Report LBL-8204, 1978.
12. McFarland, A.R., C.A. Ortiz, and R.W. Bertch, Jr. Particle
Collection Characteristics of a Single-stage Dichotomous
Sampler. Environ. Sci. Technol. 12:679-682, 1978.
APPENDIX
The values of /St for the large particle and small particle
collection efficiency curves are tabulated in Tables A-l and
A-2, respectively. The cases are the same as those listed in
Table 1. The base case is presented first and subsequent cases
are labeled by the value of the changed variable. For example,
for the case Re = 1, all variables except Re are at the base
values.
The large particle collection efficiency curves are pre-
sented in the appropriate figures. The small particle collection
efficiency curves are not presented, but were used to determine
the collection probe loss curves as described in Figure 4. The
collection probe loss curves are then presented in this paper.
51
-------
TABLE A-l. VALUES OF /St FOR THE LARGE PARTICLE COLLECTION
EFFICIENCY CURVES
Case 16%
Base
Re = 1
10
100
500
1,000
15,000
Qj/Qo = 0.05 0.57
.15
.25
Di/Do = 1.16
1.49 .37
Lo/Do = 0.013
S/Do =0.25
2
Go = 30°
Collection Probes a
(B)
(C)
(D)
Large particle
23% 35% 50%
0.48 0.59
.42 .53
.40 .51
.54 .67
.56 .70
.51 .64
.48 .59
.73
.34 .49
.29
.57
.51 .62
.48 .58
.47 .58
.49 .60
.65
.57
.59
.59
0.67
.62
.60
.76
.80
.73
.67
.79
.59
.45
.71
.67
.65
.69
collection efficiency
60% 70% 90% 98%
0.71
.67
.66
.82
.85
.77
.71
0.85
.64
.57
.69
.80
.73
.78
.69
.71
.71
0.84
.88
.86
1.02
1.03
.94
.83
.93
.78
.69
.80
.89
.85
.87
.87
.96
.80
.83
.84
1.01
1.22
1.34
1.41
1.23
1.00
1.11
.96
.86
1.13
1.06
.98
. 1.07
a Collection probe A - thin wall
Collection probe B - infinite wall thickness
Collection probe C - finite wall thickness
Collection probe D - finite wall thickness with taper
(Collection probe designs are shown in Figure 18.)
52
-------
TABLE A-2. VALUES OF /St FOR THE SMALL PARTICLE COLLECTION
Small particle
Case 84% 77% 65%
Base
Re = 1
10
100
500
1,000
15,000
QX/QO = 0.05 0.23
.15
.25
DX/DQ = 1.16
1.49
LQ/DO = 0.013
S/DQ = 0.25
2
e0 = 30°
Collection Probes3
(B)
(C)
(D)
0.23 0.39
.35 .46
.35 .47
.41 .57
.34 .54
.28 .45
.24 .39
.46
.32
.08
.40
.23 .43
.23 .38
.10 .31
.29 .44
.43
.34
.45
collection
50% 40%
0.47
.55
.56
.68
.66
.55
.47
.53
.43
.32
.52
.46
.40
.53
0.51
.60
.61
.75
.72
.59
.51
.48
.39
.51
.55
.47
.57
efficiency
30% 10%
0.58
.79
.79
.96
.92
.71
.58
0.58 .60
.56
.45 .53
.59
.61 .67
.53 .56
.46 .49
.60 .64
.62
.54
.65
2%
0.64
1.00
1.23
1.13
.64
.66
.63
.60
.74
.63
.55
.69
a Collection probe A - thin wall
Collection probe B - infinite wall thickness
Collection probe C - finite wall thickness
Collection probe D - finite wall thickness with taper
(Collection probe designs are shown in Figure 18.)
53
-------
PAPER 3
A HEAVY GRAIN-LOADING IMPACTOR
DALE A. LUNDGREN
ERNEST R. CERINI
UNIVERSITY OF FLORIDA
DEPARTMENT OF ENVIRONMENTAL ENGINEERING SCIENCES
AND
MICHAEL L. SMITH
ANDERSEN SAMPLERS, INC.
ABSTRACT
This paper discusses a heavy grain-loading impactor (Ander-
sen Model HCSS) which was designed for use in high grain loading
applications, where standard impactors cannot be used because
of collection surface overloading. The HCSS consists of two
impaction chambers followed by a cyclone and glass fiber thimble.
Construction of stainless steel allows its use in the high tem-
perature, corrosive gas atmosphere encountered in industrial
stack gas sampling.
Descriptive information on the HCSS is presented together
with particle size collection characteristics. Data on the mass
loading characteristics for a mineral dust aerosol are presented.
Application areas are cited.
INTRODUCTION
The Andersen Model HCSS is a special application in-stack
size-fractionating sampler developed specifically for particle
size distribution measurement in high aerosol mass concentration
gas streams (2 to 200 grams/m3 or 1 to 100 grains/ft3). Other
commercially available in-stack impactors typically have mass
deposit limitations on the order of 1 to 10 milligrams per stage,
for mineral type aerosols, to avoid particle reentrainment prob-
lems. Therefore, these conventional impactors, when operating
at 14 jlpm (0.5 cfm) flow rates for 1 to 100 minute sampling times,
are best suited to sampling in aerosol concentrations ranging
54
-------
from 0.002 to 2 grams/m3 (0.001 to 1 grain/ft3), giving total
particulate matter collections of 3.5 to 35 mg. In this new
sampler the particle fractionating stage surfaces are not coated
with grease or filter media, making the recovery of collected
material easier and chemical analysis simpler because analysis
coating reactions and analysis interferences are eliminated.
Because of stainless steel construction, sampling at high tempera-
tures is possible (a unit has been successfully field tested
at 1520°F) without the particle reentrainment problem often en-
countered when conventional impactors operate at high tempera-
tures without collection surface coatings. This unit will fit
through a 3-inch or larger sampling port and will operate in
both the vertical and horizontal positions. The long length
of the HCSS makes it less convenient to use in small diameter
stacks and the difficulty of accurate recovery of small mass
quantities makes it undesirable for use when sampling gas streams
with aerosol concentrations less than 0.2 grams/m3 (0.1 grain/ft3)
Sampling of. liquid aerosols is also possible if the final
filter does not become overloaded. The glass fiber thimble pro-
vides a very large collection surface area; therefore overloading
is unlikely with most aerosols. When sampling in very high tem-
perature environments the glass fiber thimble is replaced by
an alundum thimble. Because only three size fractionating stages
are used, with a last stage classification at 1.5 jam (or optional
2.5 ym), the HCSS is not superior to, nor does it replace, con-
ventional impactors. As stated, the HCSS is a special applica-
tion size fractionating sampler, capable of giving good particle
size distribution data in high aerosol mass concentration gas
streams (where conventional impactors will overload and produce
incorrect particle size distribution data).
SAMPLER DESCRIPTION
The Andersen HCSS consists of two impaction stages followed
by a cyclone stage and a glass fiber thimble filter. This unit
is shown schematically in Figure 1.
In use, an appropriate nozzle size is selected to enable
isokinet.ic sampling within the approximate flow rate range from
8 to 20 S,pm (0.3 to 0.7 cfm) . The first impactor stage has a
nominal cutpoint of 11 ym aerodynamic diameter at a 14 Jlpm (0.5
cfm) flow rate. The aerosol passing Stage 1 flows out through
a set of three outlet tubes (only one of which is shown in Fig-
ure 1) and into the second stage impaction nozzle. Stage 2 is
similar in design to Stage 1 but has a nominal cutpoint of 6 ym
at the design flow rate. Aerosol exiting Stage 2, through a
set of three equally spaced outlet tubes, next passes through
a small cyclone of the Southern Research Institute design.l A
standard 1.5 ym cutpoint cyclone is used, with a 2.5 ym cyclone
available as an option. A high-efficiency glass fiber thimble
filter removes all remaining particle matter.
55
-------
FLOW
ACCELERATION
JET
r 0
- 5
L10 cm
SCALE
VENT
TUBE
ISOKINETIC PROBE
FIRST IMPACTION STAGE
SECOND IMPACTION STAGE
CYCuONE STAGE
GLASS FIBER
THIMBLE FILTER
Figure 1. Schematic of the Andersen Model HCSS High Grain-Loading
Impactor (from McFarland3).
56
-------
Both impaction stages are scaled versions of the Andersen
preseparators shown in detail in Figure 2. The model HCSS first
stage has the same internal dimensions and collection character-
istics as the previously tested preseparator.2 A particle size-
collection efficiency curve for a 14.2 &pm (0.5 cfm) flow rate,
from McFarland, Ortiz and Bertch,2 is reproduced as Figure 3.
Extensive tests were run on this first stage separator using
a monodisperse liquid oleic acid aerosol and a polydisperse dry
fly ash aerosol. Data on the effect of flow rate upon outpoint
and mass loading characteristics are shown in Figures 4 and 5
respectively (from McFarland, Ortiz and Bertch2).
Stage 2 is similar to Stage 1 except for the jet dimensions.
This modification changes the cutpoint as shown in unpublished
data of McFarland3 who determined the cutpoint of both stages
and both optional cyclones, as shown in Figure 6. Loading charac-
teristics of Stage 2 and the Southern Research Institute design
cyclone were not previously reported; therefore the following
performance tests were run.
FLOW
ISOKINETIC PROBE
ACCELERATION
NOZZLE
VENT
TUBE
Figure 2. Preseparator design details (from McFarland, Ortiz, and Bertch?).
57
-------
100
§ 80
i
60
UJ
o
I 40
8 20
O MONODISPERSE OIL
DROPLETS
D FLY ASH
I
I
I
5 10 20 30
AERODYNAMIC PARTICLE DIAMETER, Dp, /an
I
40 50
Figure 3. Preseparator particle size versus collection efficiency (from McFarland,
Ortiz, and Bertch2).
58
-------
20
15
£
a.
Q.
a
N
co
t-
o
a.
I-
u
10
I
I
10 20
FLOW RATE, q, Sjmm
30
40
50
Figure 4. Effect of flow rate upon preseparator outpoint (from McFarland, Ortiz,
and Bertch?).
PERFORMANCE TESTS
The loading characteristics of the HCSS second stage (6 ym
cut-size impactor) and third stage (1.5 ym cut-size cyclone)
were evaluated using dry mineral dusts: AC Fine Test Dust and
AC Coarse Test Dust (sometimes referred to as Arizona Road Dust).
These are standard polydisperse mineral dusts of fairly well-
known size distribution. Both dusts were dispersed into dry
air with a Wright Dust Feed Mechanism. "* Size distribution and
concentration of the generated aerosol were determined by means
of a standard multi-stage impactor using grease coated stainless
steel collection surfaces conditioned at 200°F for one hour.
Sampling time and flow rate were selected to produce near optimum
stage collection deposits.
After the test aerosol concentration and size distribution
were determined, the aerosol was sampled through the first two
(impactor) stages of a clean HCSS and the penetrating aerosol
concentration and size distribution were again determined using
a standard multi-stage impactor. This initial two-minute aerosol
penetration test was run on the clean HCSS sampler operating
59
-------
100
~ 80
-------
100
„ 80
c
O
01
O
ai
Z
O
o
LU
o
O
60
40
201
\ I I I I I I
0.7 1 3 5 7 10
AERODYNAMIC PARTICLE DIAMETEiR, Dp,
30
Figure 6. Particle size versus collection efficiency data for all stages of the
Andersen Model HCSS sampler (from McFarland3).
on log normal paper, are shown for both the clean HCSS (noted
as initial tests) and the dust-loaded HCSS (noted as final tests)
in Figure 7 for the AC Fine Test Dust and in Figure 8 for the
AC Coarse Test Dust. Actual dust quantities collected by each
stage of the Model HCSS are also noted. Percent of dust pene-
trating the second and third stage, based on the initial two-
minute test, the final two-minute test and the overall average
for the 60- to 100-minute loading test did not significantly
change for either stage.
Size classified mass fractions collected during the standard
multi-stage impactor test using AC Fine Test Dust were used to
determine the particle size/collection efficiency of both the
second (impactor) and third (cyclone) stages of the HCSS. Mass
concentration, before and after each respective stage, was ratioed
for each particle size fraction (to determine the fractional
penetration) and then used to calculate the stage fractional
efficiency. Percent efficiency versus particle size (as particle
61
-------
A.C. FINE TEST DUST
O D INITIAL TESTS
AC FINAL TESTS
X HCSS DATA POINTS
SIZE DISTRIBUTION
AFTER SECOND STAGE
SIZE DISTRIBUTION
AFTER CYCLONE
TEST AEROSOL
SIZE DISTRIBUTION
HCSS DUST LOADING, grams
STAGE 1
STAGE 2
CYCLONE
FILTER
1 10
AERODYNAMIC PARTICLE DIAMETER, Dp, nm
Figure 7. Inlet and outlet size distribution measurements using AC fine
test dust.
100
62
-------
99.9
99
a
a
cs>
CO
til
.90
70
50
o 30
10
0.1
\ I I I I I | I I
A. C. COARSE TEST DUST
O A* INITIAL TESTS
V + 0 FINAL TESTS
X HCSS DATA POINTS
SIZE DISTRIBUTION
AFTER CYCLONE
I I i ITl
SIZE DISTRIBUTION
AFTER SECOND STAGE
TEST AEROSOL
SIZE DISTRIBUTION
HCSS DUST LOADING, grams _
STAGE 1
STAGE 2
CYCLONE
FILTER
TOTAL
I I I i i
I I I I I I I
2.747
0.244
0.213
0.023
3.227
I I I I I I
1 10
AERODYNAMIC PARTICLE DIAMETER, Dp, urn
100
Figure 8. Inlet and outlet size distribution measurements using AC coarse
test dust.
63
-------
aerodynamic diameter) is shown in Figure 9. Monodisperse aerosol
test data obtained by McFarland3 is also shown. Similar data
for the AC Coarse Test Dust is shown in Figure 10.
The explanation for the apparent difference in the cyclone
particle size/collection efficiency curves for the two test dusts
is poor experimental accuracy due to low stage weight gain.
As many researchers have found, determining particle size/ef-
ficiency data using impactor classified particle size fractions,
obtained before and after a particle collection device, is very
difficult and serves to justify the need to run tests using known
diameter monodisperse aerosols.
SUMMARY AND CONCLUSIONS
The Andersen Model HCSS high grain-loading impactor was
tested with polydisperse mineral dust aerosols. If reentrain-
ment of dust had occurred after the accumulation of a significant
quantity of dust by a collection stage, the stage collection
efficiency would decrease and the amount of coarse particles
penetrating that stage would significantly increase. Size dis-
tribution and concentration measurements after a clean and dust-
loaded HCSS impactor indicated no measurable reentrainment of
dust. These tests were run with the HCSS impactor in a hori-
zontal position operating at a flow rate of 14.2 &pm (0.5 cfm),
IUU
*, 80
c
0)
OJ
a.
>"
i- 60
Z
UJ
o
E
U.
QJ
z
o
h-
O
UJ
Ij 20
0
U
I I
—
—
—
_ .
I I
0.1
I I I
I I I
III Ur I I"! I I&T ".— I B
I
/
/ /
J /
/
x
A
CYCLONE ~
EFFICIENCY I STAGE 2
T EFFICIENCY
/
/
/
A
/ / I —
/ /
jP //
III I I I I I I I I I I
1.0 10 30
AERODYNAMIC PARTICLE DIAMETER, Dp, ,um
Figure 9. Particle size versus collection efficiency calculated from conventional
impactor measurements using AC fine test dust (from McFarland^).
64
-------
100
I 80
V
a
u.
ID
60
O 40
u
LU
O
o
20
I I
CYCLONE
EFFICIENCY
STAGE 2
EFFICIENCY
I
I I I II I
0.1
1.0 10
AERODYNAMIC PARTICLE DIAMETER, Dp, jum
30
Figure 10. Particle size versus collection efficiency calculated from conventional
impactor measurements using AC coarse test dust (from McFarland^).
sampling isokinetically through a straight sampling inlet. This
is considered to be a realistic test and certainly indicates
that the unit performed as intended. Aerosol size distribution
plots determined using the HCSS collected particulate matter
were essentially the same as determined by the standard multi-
stage impactor, which used grease coated collection surfaces (see
Figures 7 and 8). Tests run with the unit in a vertical position
and with downward flow gave results comparable to the horizontal
position. Tests run with upward flow through a vertical unit
show somewhat higher measured penetrations.
The HCSS is recommended for aerosol size distribution mea-
surement if the aerosol mass concentration is in the range from
2 to 200 grams/m3 (1 to 100 grains/ft3); with a gas stream of
almost any gas composition compatible with type 316 stainless
steel; at any gas temperature up to 1500°F; and at any reasonable
gas pressure. For long time sampling (2 to 24 hours) the HCSS
can be used for sampling aerosol mass concentrations as low as
0.2 grams/m3 (0.1 grains/ft3).
Particle reentrainment may be a problem for certain aerosols
at flow rates of 28 £pm (1 cfm) or higher, although field tests
with a non-bouncy aerosol did not indicate a problem. McFarland,
Ortiz and Bertch reported that gravitational settling within
the HCSS first stage will cause a shift in collection character-
istics at flow rates below 7 &pm (0.25 cfm). The unit may be
65
-------
useful for sampling at these lower flow rates if the stage collec-
tion characteristics are determined.
The HCSS has the ability to fill an important gap in the
size selective sampling of aerosols: that of sampling high aerosol
mass concentration streams. Inlets to particulate collection
devices, fugitive dust sources, pneumatic conveying lines, parti-
culate process streams, long time sampling applications, or use
as a particle size classifier are example areas for application
of the HCSS.
NOTE
Since this report was prepared, several additional field
tests have been run on a very high concentration, large particle
size (si to 500 pm diameter) industrial aerosol. The larger
particles were sintered spheres with minimum adhesion properties.
Extremely high mass loadings in the first impactor stage were
obtained and a carryover of about 7% of the first stage catch
to the second stage was determined by a subsequent analysis. The
carryover was fairly constant with first stage loading (from 5
to 25 grams) and particle size (over the 10 to 500 ym size range).
This is an extreme sampling condition and represents a worst
case example. Because ^80% of the aerosol was caught on the
first stage, a 7% carryover to the second stage about doubled
the second stage catch and would have caused a significant error.
Some carryover from the second impactor stage was also determined
but was not considered significant. If appreciable quantities of
100+ um particles of a solid, spherical shape exist in the sampled
gas stream, it is recommended that the second stage catch be checked
for reentrainment and a suitable correction made.
REFERENCES
1. Smith, W.B., and R.R. Wilson, Jr. Development and Laboratory
Evaluation of a Five-Stage Cyclone System. EPA-600/7-78-008,
U.S. Environmental Protection Agency, Research Triangle
Park, NC, 1978.
2. McFarland, A.R., C.A. Ortiz, and R.W. Bertch, Jr. A High
Capacity Preseparator for Collecting Large Particles. Atmos.
Environ. 13:761, 1979.
3. McFarland, A.R. Laboratory Evaluation of Andersen Hi-Capacity
Stack Sampler. Unpublished engineering report prepared
for Andersen Samplers, Inc., December, 1978.
4. Wright, B.M. A New Dust Feed Mechanism. J. Sci. Instrum.
27:12, 1950.
66
-------
PAPER 4
VELOCIMETRIC DETERMINATION OF AERODYNAMIC DIAMETER
IN THE RANGE FROM 0.1 ym TO 15 ym
JAMES C. WILSON
BENJAMIN Y.H. LIU
PARTICLE TECHNOLOGY LABORATORY
MECHANICAL ENGINEERING DEPARTMENT
UNIVERSITY OF MINNESOTA
ABSTRACT
A technique for determining the aerodynamic diameter of
aerosol particles is described. Particles are accelerated in
a nozzle, and their velocity is measured near the nozzle exit.
Careful choice of nozzle geometry and flow rate permits aerodynamic
diameter to be determined from the particle velocity.
Experimental tests are reported in which velocity is mea-
sured with a laser-Doppler velocimeter. Theoretical analysis
of the experimental tests shows that particle velocity can be
accurately predicted.
Theoretical study of particle dynamics in the nozzle shows
that both aerodynamic diameter and density affect particle velo-
city if the motion is ultra-Stokesian. This introduces uncer-
tainty in determination of aerodynamic diameter for such par-
ticles and provides an incentive for minimizing the particle
Reynolds number.
Nozzle configurations and flow rates are proposed which
should permit determination of aerodynamic diameter in the fol-
lowing ranges: 0.1 ym to 1.5 ym, 0.5 ym to 10 ym, 1.5 ym to
15 ym, and 1 ym to 15 ym.
NOMENCLATURE
a Radius of the nozzle at the exit
C Slip correction
C Drag coefficient
d Distance from nozzle exit to point of measurement
67
-------
d_ Fringe spacing
D Aerodynamic diameter
a
D Particle diameter
P
AD Diameter interval
f Doppler frequency
F, Body force acting on a particle
F, Drag force acting on a particle
g Gravitational acceleration
m Particle mass
P Pressure upstream of the nozzle
AP Pressure drop across the nozzle
r Radius of trajectory of particle executing curved motion
R Radius of nozzle at position x
Re Reynolds number, based on D
Re Particle Reynolds number
St Stokes number
t Time
U Gas velocity at exit of the nozzle
U Gas velocity
9
U * Dimensionless gas velocity
U Particle velocity
U * Dimensionless particle velocity
U Geometric mean particle velocity
pg
U Terminal velocity of particle settling under gravity
x Position along centerline of nozzle axis
a Angle of convergence of the nozzle
68
-------
X Wavelength of laser radiation
y Viscosity of gas
p Density of the gas
p Density of the particle
o Geometric standard deviation of size distribution
0 Geometric standard deviation of velocity distribution
T Period of Doppler signal
w Angular velocity of a particle in curved motion
INTRODUCTION
Aerodynamic diameter is an important measure of the size
of aerosol particles. It is a single parameter combining par-
ticle density, diameter, shape factor, and slip correction, and
is used to predict the motion of particles in settling, impac-
tion, or in a centrifugal force field. These motions frequently
cause particle deposition in the respiratory tract and particle
collection in gas cleaning devices such as cyclones and settling
chambers. Consequently, it is often desirable to make accurate
measurements of aerodynamic diameter in order to study and pre-
dict these phenomena.
The popularity of cascade impactors, cyclones, and impingers
is due in part to the fact that they measure aerodynamic diam-
eter. However, these devices separate the particles from the
suspending gas, and their use often involves time-consuming
analysis of the collected particles.
This paper describes a new instrument which allows rapid,
in-situ measurement of aerodynamic diameter. The instrument
makes use of a nozzle to accelerate the particles and a laser-
Doppler velocimeter to measure the particle velocity near the
nozzle exit, as shown in Figure 1.
Small particles are able to follow the flow as it accele-
rates rapidly in the nozzle. Such particles emerge from the
nozzle with velocities near that of the gas. Large particles
lag behind the gas flow and emerge with lower velocities. The
velocity of individual particles emerging from the nozzle can
be measured by a laser-Doppler velocimeter. In this system,
the laser beam is split and the two coherent beams are focused
by the transmitting optics to a crossover point where they form
interference fringes. A particle passing through the fringes
scatters light, and the scattered intensity rises and falls as
69
-------
AEROSOL
,CLEAN AIR
LASER
BEAM
SPLITTER
PHOTOMULTIPLIER
WITH PREAMP
OSCILLOSCOPE
Figure 1. Schematic of the laser-Doppler velocimeter system.
the particle passes through the bright and dark bands. The
scattered light is collected and focused on a photomultiplier
tube (PMT), and the frequency of the intensity modulation is
counted. The particle velocity perpendicular to the fringes
equals this Doppler frequency, fv, multiplied by the fringe
spacing, df, which is calculated from the wavelength of the laser
radiation and the angle between the beams. By choosing appro-
priate nozzle parameters and flow rates, the aerodynamic diameter
can be determined from the measured particle velocity.
Other methods of determining particle size from particle
velocity are reviewed elsewhere.
THEORETICAL CONSIDERATIONS
Aerodynamic Diameter
The definition of aerodynamic diameter and its role in de-
termining particle motion strongly affect the design of this
instrument. Aerodynamic diameter is often defined as the diam-
eter of a unit density sphere with the same settling velocity
as the particle in question.2 The usual mathematical expression
for aerodynamic diameter is derived here by considering the
motion of particles which settle with small values of the par-
ticle Reynolds number.
Equation 1 is the equation of particle motion and Equation 2
is Stokes law.
aS ^
•at*'*
(1)
Fd =
(0
q
- u )
p
3
TT
P
D
P
(2)
70
-------
Stokes law accurately describes the drag force on a spherical
particle moving with a Reynolds number, Rep, less than about
0.5, where
|U - U I p D
- " P
In the case of a particle settling in still air under gravity,
Ua = 0 and Fu, = mg. The terminal velocity, Ut, is reached when
tne drag force equals the weight. Thus, for a particle settling
in Stokes regime (i.e., Re <0.5),
C D 2 p g
ut = —fe—
Since D , the aerodynamic diameter, is the diameter of the unit
density sphere having the same Ut as the particle in question,
D_ can be calculated as follows:
Cl
)Pp] Dp . (5)
Non-spherical particles are treated by including the dynamic
shape factor.3' **
Equation 5 does not follow from the settling velocity defini-
tion if the settling occurs outside of Stokes regime, as is often
the case for large particles. However, even in such cases, Equa-
tion 5 is taken as the definition of aerodynamic diameter because
Dg is useful in predicting other motions of the particle which
may occur in Stokes regime. For example, consider impaction
(FL = 0) and centrifugal motion (F^ = mru)2) . Combining Equa-
tions 1 and 2 with the appropriate expression for Fb results
in an equation which can be written in terms of Da and solved
for the particle trajectory. Thus, for given flow field and
initial conditions, the particle trajectory in an impactor or
cyclone is determined by Da if the motion is Stokesian. This
is true whether or not the particle settles in Stokes regime.
The Present Technique
In the method described here, Da is determined by accele-
rating the particle in a converging nozzle and measuring its
velocity near the nozzle exit. Appropriate flow conditions must
be chosen if Da is to be accurately determined from the particle
velocity. Dimensional analysis of the equations of motion indi-
cates the factors involved in the choice.
71
-------
Consider a particle on the axis of the nozzle shown in Fig
ure 1. Equation 1 describes the motion of the particle where
Fb equals zero and F^ is expressed in Equation 6.
TT p (U - U ) 2 D 2 C
- ~^—a - P °
a ~ e
CD is the drag coefficient. For Rep <0.5, Stokes' law is ex-
pressed as
CD = 24/Rep . (7)
For 0.5 < ReD < 100, Fuchs3 suggests the following expression
for CD:
Equations 1 and 6 are combined and written in dimensionless form
to produce Equation 9,
dU * Re (U * - U *)2C
2 - D {9)
dx* 24(St)U *
P
where St is the Stokes number
and
St = C D 2 p U /18 M a (10)
p p e
Re = Ue Dp p/y . (11)
U is the centerline gas velocity at the nozzle exit and a is
the nozzle exit radius. Also,
U * = U /U (12)
g g e
V = V°e (13)
x* = x/a (14)
and
Re = Re (U * - U *) . (15)
72
-------
Substitution of Equations 7 and 15 into 9 shows that Up*
at a given point depends upon St for motion in which Rep <0.5.
Thus, for motion in Stokes regime, St, and hence, Da, can be
determined from Up*. In cases where Rep >0.5, the use of Equa-
tion 8 implies that Up* depends upon Re as well as St. Thus,
particle velocity depends upon the product of ppC, as well as
Da, when Rep is large. Therefore, particles with the same ve-
locity may nave different values of Da if they have different
densities. This introduces an uncertainty in the determination
of Da from Up* when p1 is unknown and Rep is large. This un-
certainty is minimized in this method by minimizing gas velocity,
and hence, Rep. This distinguishes the present method from those
which employ nozzles operating at large values of pressure drop.
EXPERIMENTAL TEST
Measurements were made of the velocity of particles emerging
from a nozzle. Particle size and flow through the nozzle were
varied. The experimental system is diagrammed in Figure 2.
The actual test nozzle was 1.77 cm in length and had entrance
and exit diameters of 1.05 cm and 0.106 cm, respectively. A
short throat was found near the nozzle exit. The pressure across
the nozzle was varied between 2.54 cm H20 and 691 cm H20. Sheath
air was used to confine the aerosol near to the center stream-
line.
Monodisperse aerosols were generated from PSL and PVT sus-
pensions, as well as with a vibrating orifice aerosol generator.
The velocity of individual particles was measured with a laser-
Doppler velocimeter using a He-Ne laser. A fringe spacing of
SIGNAL PROCESSING
_L
AEROSOL
GENERATOR
OL
1
INLET
PRESSURE
11
V 1
U
1
H
1
| PHOTOMULTIPLIER
o
1 r
M
^h
CHAMBER
(?) PRESSURE
' *• PUMP
COMPRESSED AIR
FILTER
AEROSOL'
FLOWMETER
COMPRESSED
AIR
, BEAM
I SPLITTER
LASER
SHEATH AIR
FLOWMETER
Figure 2. Experimental system for the test nozzle.
73
-------
15.9 ym was used for particles larger than 3 ym, and a 15 mw
He-Ne laser and 7.13 ym fringe spacing was used for smaller par-
ticles. The center of the measuring volume was approximately
145 ym from the exit of the nozzle and extended four fringes
on either side of this point. Details of the experimental pro-
cedure and results are described in detail elsewhere.1'5
Each experimental trial resulted in a frequency distribution
of velocities which is characterized by a geometric mean velocity,
Upg, and a geometric standard deviation, agv. The geometric
mean velocities are plotted as a function of particle size and
for two values of nozzle pressure drop in Figure 3.
The resolution which could be expected from this system
can be estimated from the values of agv and the particle velocity
vs. diameter curve. 6
o
i-
cr
&
\
X
EK
B
cm of H^O
""-•069.1
£E3*
0 I 23456789 10 II 12
PARTICLE DIAMETER , p.m
Figure 3. Experimental and theoretical particle velocity at a distance of
145 urn from the nozzle exit as a function of particle diameter
and pressure drop across the test nozzle.
74
-------
TABLE 1. SELECTED VALUES OF agv AND CORRESPONDING DIAMETER
INTERVALS FOR THE TEST NOZZLE
AP = 25.4 cm H20 AP = 69.1 cm H20
Dp, ym agv ADp, Vm °gv ADp, ym
0.
3.
9
5
04
1.
1.
1.
007
005
007
±0.
±0.
±0.
25
05
15
1.
1.
1.
02
006
008
±0.
±0.
±0.
19
06
13
If the test aerosols were truly monodisperse, ADp would provide
a good estimate of the resolution of the system. In the present
case, the system resolution for larger particles is smaller than
ADp, since the actual width of the test aerosol size distribution
is significant compared to ADp. For the submicron particles,
the slope of the velocity vs. diameter curve limits resolution
for small values of pressure drop, and the increase in agv, per-
haps due to turbulence, limits resolution at large values of
pressure drop.
TEST OF THE THEORY
Theoretical calculations were made of particle velocity
in the test nozzle. These values were compared with the experi-
mental results as a check on the theory. The first step in the
calculation was to calculate the air velocity in the nozzle from
the pressure drop and the nozzle dimensions. Some asymmetries
in the nozzle shape required that average values of radius be
determined at positions along the axis to allow an axisymmetric
representation of the nozzle. The axisymmetric representation
of the nozzle exit region is outlined in Figure 4.
The velocity of the air in the nozzle was calculated using
two methods. Boundary layer calculations6'7 were used to deter-
mine the velocity at every point in the nozzle, and Bernoulli's
law applied to plug flow was used to calculate only centerline
velocities. Both calculations were made for the constant prop-
erty fluid case. (See reference 5 for details.) The two methods
agreed quite well on the centerline velocity, as is shown in
Figure 4. The boundary layer calculations also provided informa-
tion on the flow profile in the test nozzle. The calculated
flow profiles suggest that even for the lowest pressure drop,
a sheath air flow equal to one-half of the total flow is adequate
to insure that all particles experience nearly the same flow
field. Therefore, only the centerline velocities are used in
the calculations of particle velocity.
75
-------
PLUG FLOW
CALCULATION
BOUNDARY
LAYER AP, cm of H_0
CALCULATION
0.8 1.0 1.2 1.4
POSITION ALONG AXIS , cm
Figure 4. Air velocity on the centerline of the test nozzle calculated using
the boundary-layer approximation and Bernoulli's law applied to plug
flow. The axisymmetric representation of the nozzle exit region is
also shown.
The particle velocity was calculated using a Runge-Kutta
technique to solve Equation 1 with the drag force described by
Equation 6 and the drag coefficients described by Equations 7
and 8. The particle velocity measurements were made approxi-
mately 145 urn downstream of the nozzle exit; therefore, the
integration was continued to this point. The gas velocity was
assumed constant over this distance. The particle and gas ve-
locities were assumed identical at the entrance to the nozzle.
The results of the calculations are shown in Figure 3.
Comparison of the experimental and theoretical values of
particle velocity shows good agreement. The percent mean devia-
tion between the measured and calculated values is on the order
of 1.1% for AP = 61.9 cm of H20. At a AP = 276 cm of H20, the
incompressible flow assumption should begin to fail, yet the
mean deviation between measured and calculated velocities is
only about 3%.
76
-------
MEASUREMENT OF AERODYNAMIC DIAMETER
Measurement of aerodynamic diameter by this technique re-
quires careful choice of nozzle and flow parameters. These
parameters determine the resolution of the systems (particularly
for smaller particles), the uncertainty introduced by uncertain
density and the aerosol sample rate. General solutions to the
dimensionless equation of motion were obtained to facilitate
these choices. Figure 5 shows the generalized nozzle geometry
used in the investigation. Three angles, a = 15°, 30°, and 45°,
were selected, and the dimensionless gas velocity on the center-
line of each nozzle was calculated using Bernoulli's law for
ideal, one-dimensional flow. The gas velocity was assumed to
be constant over the distance d from the nozzle exit to the point
of measurement.
Equation 9 was then solved numerically to find the dimen-
sionless particle velocity, Up*, for each nozzle as a function
of St, Re, and d. Figure 6 shows results for a = 45° and
d = 0.2a. Solution sets were obtained for ;, = 15°, 30", and
45°, with d = 0, 0.2a, 0.4a, and 0.6a.5 In the case shown in
Figure 6, the principal change in dimensionless velocity occurs
between (St)% = 0.4 and (St)% = 4. This geometry, then, should
allow approximately one decade in Da to be measured for a given
flow rate. To determine a calibration curve, the smallest aero-
dynamic diameter to be measured with this system is assigned
a value of (St)35 equal to about 0.4. This fixes the ratio of
Ue/a. Choosing the total flow in the nozzle then determines
Ue and a. The curves are calculated for incompressible flow
in the nozzle; therefore, the pressure drop across the nozzle
must be small compared to the upstream pressure. Once Ue and
a are chosen, values of St, Re, and Da are determined from
Equations 10, 11, and 5 for particles with the desired densities.
Then Up* is determined from the curves in Figure 6 and is plotted
against Da.
tana
Figure 5. Nozzle shape and parameters used in the theoretical analysis.
77
-------
0.2 0.30.4 0.60,81.0 2.0 3.04.0 6.08.010 15
Figure 6. Dimension/ess particle velocity as a function of Re and (St)^2 for a
nozzle having a = 450 and d =* 0.2a.
Figures 7 and 8 show curves of UD* vs.
"P •! f* 1 1 T" O f\ T VN W y^ 4- V% ^»t -^^^^N<^ /^i — A d O
Da obtained from
In both cases, a = 45°, a - 0.05 cm, and d/a = 0.2.
However, Ue = 9500 cm/s in the first case, and Ue = 1200 cm/s
in the second.
The system shown in Figure 1 should permit the measurement
of Da for atmospheric aerosols in the range from 0.5 ym to 10 ynu
For particles having Da larger than about 2 pm, an uncertainty
is introduced into the determination of Da from Up* if particle
density is unknown. This uncertainty increases with Da and can
be reduced by reducing Ue, as is done in Figure 8. Here, resolu-
tion is lost for submicron particles, and the sample rate is
reduced to a level unsuitable for atmospheric sampling; however,
an even wider range of densities results in a smaller uncertainty
in Da. Experimental tests are required to determine the resolu-
tion for small particles and the maximum suitable sample flow
rate. However, sample flow rates near 37 cm3/s and 4.7 cm3/s,
respectively, are expected for the two cases.
Different nozzle and flow parameters result in the curves
shown in Figure 9. The nozzle exit diameter is sufficiently
large so that velocity measurement can take place very near the
exit, although the value of d = 0 is an exaggeration. Reducing
d/a moves the curves in Figure 6 to the left. The aerosol sample
flow rate in this case is estimated to be 500 cm3/s and should
be sufficient to permit sampling of coarse particles in the at-
mosphere. The uncertainty introduced by density variations be-
tween 1 and 4.5 g/cm3 is limited to about ±5% if p = 2 g/cm3
is chosen as the reference curve.
78
-------
0=0.05 cm
d=0.01 cm
Ue = 9500cm/s
P= I atm
23456789
AERODYNAMIC DIAMETER ,
Figure 7. Calibration curve for a nozzle whose parameters were chosen to permit
measurement of atmospheric aerosols in the 0.5 urn to 10 nm range.
The sample flow rate would be approximately 37 cm^/s.
i.o
o
>"
K
O
8
0.9
2 0.8
UJ
y 07
H
CC
3.
0.6
V)
u
O
V)
05
0.4-
a3
a = 45°
o = 0.05cm
d=0.01 cm
Ue= I200cm/s
P=lotm
= 2g/cm3 -
I I I I
0 I 2 3 4 5 6 7 8 9 10 II 12 13 14 15 16 17
AERODYNAMIC DIAMETER , ^m
Figure 8. Calibration curve for a nozzle whose parameters were chosen to permit
measurement of aerosols in the 1.5 pm to 15 nm range. The sample
flow rate would be approximately 4.7 cm^/s.
79
-------
1.0
*^09
tOfl
3
UJ
> 0.7
UJ
o
H C
tr
Q.
,0 0.5
UJ
_J
§0.4
so.:
o
0.2
-i—i—i—r
o=. 0.29 cm
d=0 cm
3870cm/s
1atm
I 2 3 4 5 6 7 8 9 10 II 12 13 14 15 16
AERODYNAMIC DIAMETER,^m
Figure 9. Calibration curve for a nozzle whose parameters were chosen to permit
measurement of particles in the range from 1.5 JU/T? to 15 pm. The
sample flow rate should be adequate to permit convenient sampling of
coarse particles in the atmosphere.
Measurement of Da between 0.1 and 1.5 ym may be accomplished
with a nozzle operating at reduced pressure. In the case shown
in Figure 10, the pressure upstream of the nozzle is assumed
to be 0.5 atm. This may be accomplished by passing the aerosol
through a critical orifice upstream of the nozzle. The pressure
drop in the nozzle itself is about 0.16 of the upstream pressure.
The values of aerodynamic diameter shown in Figure 10 were cal-
culated for 1 atm pressure. Experimental tests will be required
to determine the effect of turbulence on the resolution and the
suitable aerosol sample flow rate. Aerodynamic diameter is not
often useful in predicting the motion of particles having diam-
eters of a few tenths of a micron. For such particles, diffusion
is often more important than impaction. However, measurements
of Da may be as informative in this size range as the conventional
measurements of optical diameter.
CONCLUSIONS
A velocimetric technique for measuring aerodynamic diameter
has been described. The method involves accelerating particles
in a nozzle and measuring their velocity near the nozzle exit.
Nozzle geometries and flow rates have been proposed for measure-
ment of aerodynamic diameter in the following ranges: 0.1 ym
to 1.5 ym, 0.5 ym to 10 ym, 1.5 ym to 15 ym, and 1 ym to 15 ym.
Uncertain particle density introduces uncertainty in the measure-
ment for the larger particles in each range, and the smallest
80
-------
1.0
*.» 0.9
5 0.8
UJ
> 0.7
£0.6
K
<
Q.
0.5
CO
in
UJ
z 0.4
O
0.3
0.2
I I I I I i I I I I i I i i i i
a = 45°
a = 0.01cm
d= 0.004 cm.
U,= l.65 X 10 cm/s
P- 0.5 atm
1g/cm3
p- 3g/cm3-
i i i i
i i i i i
/»= 2g/cm 3-
j i
.I .2 .3 4 .5 .6 .7 .8 .9 I.O I.I 12 I.3 1.4 1.5 1.6 1.7 1.8
AERODYNAMIC DIAMETER, /im AT STP
Figure 10. Calibration curve for a nozzle whose parameters were chosen to
permit measurement of particles in the 0.1 yjn to 1.5 pm range.
Pressure upstream of the nozzle is assumed to be 0.5 atm.
particles in each range may be affected by turbulence which de-
grades resolution. This approach allows aerodynamic diameter
to be accurately measured rapidly and in situ.
ACKNOWLEDGEMENTS
This work was completed under a contract, No. EY-76-S-02-
1248, from the U.S. Energy Research and Development Administra-
tion. The laser-Doppler velocimeter was loaned by TSI, Inc.,
St. Paul, Minnesota. This report is Particle Technology Labora-
tory Publication No. 398.
REFERENCES
1.
2.
Wilson, J.C., and B.Y.H. Liu. Aerodynamic Particle Size
Measurement by Laser-Doppler Velocimetry. To be published
in J. Aerosol Sci.
International Commission on Radiological Protection Task
Group on Lung Dynamics. Deposition and Retention Models
for Internal Dosimetry of Human Respiratory Tract. Health
Phys. 12:173-207, 1966.
3.
Fuchs, N.A. The Mechanics of Aerosols.
New York, 1964.
Pergamon Press,
81
-------
4. Raabe, O.G. Aerosol Aerodynamic Size Conventions for In-
ertial Sampler Calibration. J. Air Pollut. Control Assoc.
26:856-860, 1976.
5. Wilson, J.C. Aerodynamic Particle Size Measurement by Laser-
Doppler Velocimetry. Ph.D. Thesis, University of Minnesota,
Mechanical Engineering Department, Particle Technology
Laboratory, 1978.
6. Patankar, S.V., and B.D. Spaulding. Heat and Mass Transfer
in Boundary Layers, 2nd Ed. Intertext Books, London, 1970.
7. Sparrow, E.M., B.R. Baliga, and S.V. Patankar. Heat Trans-
fer and Fluid Flow Analysis of Interrupted-Wall Channels,
with Application of Heat Exchangers. J. Heat Transfer,
99 Series C:4-ll, 1977.
82
-------
PAPER 5
A PROTOTYPE PARTICULATE STACK SAMPLER
WITH SINGLE-CUT NOZZLE AND
MICROCOMPUTER CALCULATING/DISPLAY SYSTEM
JOHN C. ELDER
LARRY G. LITTLEFIELD
MARVIN I. TILLERY
LOS ALAMOS SCIENTIFIC LABORATORY
UNIVERSITY OF CALIFORNIA
ABSTRACT
A prototype particulate stack sampler (PPSS) has been de-
veloped to improve on the existing EPA Method 5 sampling apparatus,
Its primary features are (1) higher sampling rate (56 2,/min) ;
(2) display (on demand) of all required variables and calculated
values by a microcomputer-based calculating and display system;
(3) continuous stack gas moisture determination; (4) a virtual
impactor nozzle with 3 ym mass median diameter cutpoint which
collects fine and coarse particle fractions on separate glass
fiber filters; (5) a variable-area inlet to maintain isokinetic
sampling conditions; and (6) conversion to stainless steel com-
ponents from the glass specified by EPA Method 5. The basic
sampling techniques of EPA Method 5 have been retained; however,
versatility in the form of optional in-stack filters and general
modernization of the stack sampler have been provided in the
prototype design. Laboratory testing with monodisperse dye
aerosols has shown the present variable inlet, virtual impactor
nozzle to have a collection efficiency which is less than 77%
and significant wall losses. This is primarily due to lack of
symmetry in this rectangular jet impactor and short transition
lengths dictated by physical design constraints (required passage
of the nozzle through a 7.6 cm (3 in) diameter stack port).
Electronic components have shown acceptable service in laboratory
testing although no field testing of the prototype under a broad
range of temperature, humidity, and S02 concentration has been
undertaken.
83
-------
INTRODUCTION
The standard manual stack sampling method for particulates,
EPA Method 5,1 was reviewed to identify needed improvements.2
A prototype particulate stack sampler (PPSS) incorporating the
more desirable improvements into a general purpose particulate
stack sampler has been developed. Increasing the flow rate
(double the 28 &/min °f most Method 5 samplers commercially
available) permits shorter sampling time, collection of a larger
sample, or greater sensitivity. Other major improvements include:
electronic calculating/display of calculated variables such as
stack velocity, sample volume, and per cent of isokinetic sampling
conditions; electronic continuous readout instrumentation for
temperature, pressure, flow, and humidity; reduced weight in
individual packages; stainless steel surfaces contacting the
gas stream; improved structural strength to reduce breakage;
single-point particle size classification (3 pm aerodynamic
diameter) in the nozzle; and optional in-stack particulate filter
sampler to simplify sampling at low stack moisture conditions.
Development of a null-probe device that would greatly simplify
stack sampling was discontinued due to technical problems.
It was not our intention to provide an all-purpose sampler
capable of particulate sampling under all environmental condi-
tions. Design specifications that the PPSS will meet are com-
pared with typical Method 5 capabilities in Table 1. Our ex-
perience in the early stages of the study and experience of others
has shown that not all the conditions encountered in particulate
stack sampling can be accommodated by a single sampler. It was
therefore decided that the design specifications should accommo-
date the most common ranges of stack temperatures, pressures,
and humidities found in actual use. Several options could also
be provided to allow sampling under special conditions outside
these design limits. The PPSS will not, for example, be appli-
cable to high-temperature conditions in the typical incinerator
stack, or to the high-temperature, nearly saturated conditions
of the power plant stack at the outlet of a scrubber where drop-
lets interfere with particle size classification in the nozzle.
SI or metric units have been incorporated into the PPSS
to replace the British system of engineering units commonly used
in existing samplers. Consistency of units is observed within
the PPSS, and external data, such as barometric pressure, are
entered into the calculating/display system in the proper units.
PROTOTYPE PARTICULATE STACK SAMPLER (PPSS) COMPONENTS
The PPSS is shown schematically in Figure 1 and photographi-
cally in Figure 2. The primary components are (1) an in-stack
variable-inlet, virtual impactor nozzle capable of inertially
separating the particle size distribution into two fractions
84
-------
TABLE 1. EPA METHOD 5 AND THE PPSS COMPARED
Feature
Method 5
PPSS
Nozzle cutpoint
Maximum port diameter
Maximum weight/package
Sample flowrate
Material
Stack velocity measure-
ment
Maximum stack tempera-
ture with cooling
without cooling
Maximum stack diameter
Maximum stack gas
velocity
Maximum stack gas
humidity
Particulate filters
Number of samples
Calculation/display
Probe washing
None
7.6 cm (3 in.)
Approximately 27 kg
28 2,/min
Pyrex glass
S-type pitot
1000°C
320°C
9 m (30 ft)
>22 m/s
Saturated
Note b
5C
Nomograph calculation;
display by dials,
inclined gauge
Required
2.5-3.5 um
7.6 cm (3 in.)
24 kga
56 Jl/min
Stainless steel
S-type pitot
Not applicable
320°C
6 m (20 ft)
22 m/s
Nearly saturated
Note b
3C
Digital display
on demand (by
microcomputer)
Not required
after nozzle
characteris-
tics are known.
Notes
b.
c.
Production model weight can probably be reduced
approximately 20%.
Both methods can be provided with either in-stack or
out-of-stack particulate filters.
Number includes samples requiring preweighing, handling,
and analysis by some method; i.e., filter containing
solid material washed from probe, main gas filter, bleed
gas filter, impinger volume, or desiccant weight.
85
-------
oo
VARIABLE
INLET
NOZZLE
=(jT) STACK 1
> i^-PITOT A
^— zr — ' 01 F
4
to-t-
rr-n &
fEMP
R STACK AP
ED FLOW pJ
I
I
I-BEAM SUPPORT (f
J ^ PROBE
^AAAAAAAAAAAAA/V AAAAAA/^VWi
T ; L T • «•!
Y-U-T— — — ' —ir-r=.
r NOZZLE INLET PROBE =
\ POSITIONING
' ROD
I
n:
TEMPERATURE
EXTERNAL TEMPERATURE ~^\
yWVNAAAAAAA/VvWy ^A^\AAA/VSf- 1
-j- '•' '" - - - / /
Is* / EXTENSION ' Pi
^COUPLING NOZZLE INLET
POSITION INDICA
> r ~
y ^o
/
\
TOR
-—
=
A
~]S>
/FILTER
0 .-.TEMP
rf]
JT
i
BLEI
FLO\
J
B
— V
XFILTER HOLDER
STACK
FLOW
1
STACK
WALL
CONTROL MODULE
! PSYCHROMETER MODULE
„._____..,
EXTENSION
HEATER CONTROL
.FILTER
r
UMBILICAL l_
1
STACK APJ-LAP
COMP| '
r— IPITOTAP h-r
1
_ ,. ^. .
_1 __.
*
IDRYERS
Figure 1. Schematic diagram of the PPSS.
-------
00
Figure 2. Overall arrangement of the PPSS modules.
-------
(single cut-point capability), (2) a straight, heated, stainless
steel probe with smooth internal surfaces and an extension to
provide up to 3-m inside-stack length, (3) an insulated 200-mm-
diameter filter holder, (4) a wet-bulb, dry-bulb psychrometer, (5)
dii-cuoled or water-cooled desiccant dryer, (6) electronic flow
metPrs, (7) carbon vane rotary pump, and (8) electronic tempera-
ture and pressure instrumentation.
Variable-Inlet, Virtual Impactor Nozzle
The variable-inlet, virtual impactor nozzle, shown in Fig-
ure 3, was designed to pass through a 7.6 cm (3 in) stack port.
The rectangular virtual impactor was originally proposed as a
versatile single cut-point sampler by Forney.3 Virtual impac-
tion was expected to provide the advantage of low particle re-
bound and minimal wall losses. Further, the rectangular shape
of the jet was compatible with the proposed rectangular variable
area inlet.
•VIRTUAL CHAMBER
-SEAL RINGS
POSITIONING
ROD
INLET
Figure 3. Section of variable-inlet, virtual impactor nozzle (Mod. 3).
88
-------
The nozzle was designed to separate gas borne particles
as follows: the larger particles in the gas stream intrude into
a volume of relatively stagnant air and, being unable to nego-
tiate a sharp turn at that point, proceed to a collection filter
within the nozzle; the smaller particles successfully negotiate
the turn and proceed along the probe to the main sample filter.
By selection of appropriate length of the jet (L), width of the
jet (W), separation between jet and virtual surface (S), virtual
chamber width (H), main flow (Qm) and bleed flow (Qb), the nozzle
should provide separation of particles at the desired aerodynamic
diameter cutpoint. The dimensions L, W, and S and the two flow-
rates are maintained constant during a sampling run. Isokinetic
conditions are maintained by adjusting the nozzle opening N
during the run by mechanical linkage from outside the stack.
Four versions of the variable-inlet, virtual impactor nozzle
were tested by sampling monodisperse dye aerosol produced using
the Berglund-Liu vibrating orifice aerosol generator. The ver-
sion shown in Figure 3 was the final version. Although it was
adjusted to optimum ratios of S/W and H/W recommended by Forney
et al.,1* the nozzle displayed the relatively low collection ef-
ficiencies and high wall losses, as shown in Figure 4. Peak
efficiency did not exceed 77% and exhibited an even lower effi-
ciency for aerosols as large as 14 ym mass median diameter.
This efficiency is total aerosol mass collected in the virtual
chamber (thimble filter plus chamber wall losses) as a per cent
of total aerosol mass entering the nozzle. Wall losses in the
nozzle, probe, and filter holder ranged from 34-72% with par-
ticles near the cutpoint size showing the highest losses. Figure
4 shows, in less detail, the designs of the previous three vir-
tual impactor nozzles and the performance characteristics of
each of these designs. Mod 2 displayed performance character-
istics comparable to those previously noted for Mod 3. Figure 5
shows a photograph, provided by Professors Ravenhall and Forney,
detailing flow patterns within a water model geometrically similar
to our Mod 3 nozzle. Breakup of streamlines starting before
the exit of the jet illustrates nonuniformity of flow which would
promote loss to the walls, particularly at a point opposite the
small particle exit. Further work on this particular version
is not anticipated at this time. However, if additional work
is planned, it is recommended that a symmetrical nozzle (circular
or annular in shape) without the constraint of fitting through
a 7.6 cm port be considered. Larger port clearance will allow
adequate length for smooth transitions within the inlet and
nozzle and greater separation between the inlet and the flow
disturbances caused by the broad body of the nozzle. Symmetry
will allow the main flow stream to exit in all directions, there-
by eliminating crossover and end effects inherent in rectangular
impactors.
89
-------
MOD 0
MOD 1
O
90
80
70
I ' I ' 1 ' I ' I
0. ct.c _ VIRTUAL CHAMBER MASS
/0 EFF ~ MASS ENTERING INLET _
MOD 2-
E 60
LL
111
50
o
UJ
MOD 3
O
O
40
30
20
I.I.I
2 4 6 8 10 20
MASS MEDIAN DIAMETER, /j.m
MOD 2
VIRTUAL CHAMBER
SEAL RINGS
POSITIONING
ROD
MODE 3
Figure 4. Test results of virtual impactor nozzle.
90
-------
-------
Stack Velocity Measurement
Stack velocity is determined from S-type pitot differential
pressure as measured by an electronic pressure transducer with
± 25 mm H?0 range. A null-probe device (Figure 6) which provided
static pressure taps internal and external to the nozzle was
evaluated. The small differential pressure induced by velocity
imbalance was sensed by a ± 25 mm H20 differential pressure trans-
ducer and was minimized to achieve isokinetic conditions. The
difficulty with this technique is sensing the low differential
pressure required to achieve 0.9
-------
Instrumentation to Monitor Sampling Conditions
Instrumentation in the PPSS is capable of transmitting con-
tinuous voltage signals for calculating and display purposes,
which replace the mechanical and manual methods of flow, pres-
sure, and moisture measurement now part of Method 5. These
instruments were selected to perform with accuracy and precision
at least equivalent to that provided by Method 5 instrumentation.
Field testing to establish their ruggedness has not been accomp-
lished. Instrumentation channels are listed in Table 2 and dis-
cussed in the following sections.
Flow metering—Since the particulate mass concentration
is desired in terms of dry gas volume at standard temperature
and pressure, dry gas volume is measured and automatically con-
verted to standard conditions according to perfect gas laws.
The gas stream is dried and filtered prior to entering the flow
meter. Additional correction is required if constituents of
the gas change from the calibration gas or if temperature or
pressure of the gas stream varies from calibration conditions.
Gas analysis prior to each particulate sampling run is required
to provide gas constants to determine mass flow meter correction.
Two flow meter types, hot-wire and turbine, were included
in the prototype to permit evaluation under field conditions
since laboratory testing did not show a clear advantage of one
type over the other. Datametrics hot-wire mass flow meter Model
1000.5B was used to meter the bleed flow. It is compact and
lightweight and has been used successfully in other areas of
our laboratory. The instrument is supplied with linearizing
signal conditioning. A turbine flowmeter (Flow Technology Model
FTC-8) was used to meter sample flowrate. The advantage of
the turbine flowmeter over a hot-wire flow sensor is its direct
indication of flow volume in the presence of gases other than
the components of air. Its output does, however, require com-
pensation for temperature and pressure changes.
Flow control is performed by manual adjustment of needle
valves in the main flow stream and the bleed flow stream. This
adjustment is made infrequently since normal operation with the
virtual impactor nozzle requires constant flow.
Moisture measurement—The PPSS contains a wet-bulb, dry-
bulb psychrometer for measuring water vapor content of the sample
gas stream. Total moisture can also be determined by measuring
total weight change in the dryers. The volume occupied by water
vapor is required in the isokinetic calculation to correct nozzle
inlet velocity, which is volumetric flowrate of wet gas divided
by probe nozzle cross-sectional area. The psychrometer, shown
93
-------
TABLE 2. PPSS INSTRUMENTATION CHANNELS
Sensors
Stack temperature
Probe temperature
Extension temperature
Holder temperature
Dry-bulb temperature
Wet-bulb temperature
Sample flow tempera-
ture compensation
Stack velocity (pitot)
Flow meter pressure
corrected
Psychrometer pressure
corrected
Sample flow rate
Bleed flow rate
Elapsed time
Interval time
Range or Max
20-325°C
20-325°C
20-150°C
20-150°C
20-150°C
20-150°C
20-150°C
±34 mb
±340 mb
±340 mb
7-12xlO~V/s
1-3x10" V/s
0-99 min:
0-60 s
0-99 min:
0-60 s
Type
RTD3
RTD
RTD
Thermistor
Thermistor
Thermistor
Thermistor
Variable
reluctance
Variable
reluctance
Variable
reluctance
Turbine
Hot-wire
Digital
Digital
Signal
0-5 Vdc
0-5 Vdc
0-5 Vdc
0-5 Vdc
0-5 Vdc
0-5 Vdc
0-5 Vdc
±5 Vdc
±5 Vdc
±5 Vdc
0-5 Vdc
0-5 Vdc
Overall
Precision
±5°C
±5°C
±5°C
±1°C
±0.5°C
±0.5°C
±1°C
±0.5 mb
±5 mb
±5 mb
±0.2xlO~'*m3/s
±0.1xlO~V/s
Resistance temperature detector.
in Figure 7, is located immediately behind the sample filter
where temperature of the gas stream is maintained above dew point,
The wick material is polyester, which is resistant to the acids
(primarily dilute sulfuric acid) encountered in some sampling
situations. The wick is fed from a water reservoir through a
stainless steel tube approaching within 3 mm of the thermistor
thermometer from the downstream side. Covering the feed tube
with the wick adjusts the temperature of the feedwater to ap-
proximately that of the wet bulb and prevents cooling or heating
of the wet-bulb thermistor by the feedwater.
94
-------
METAL BOX
WET-BULB
THERMISTOR
DISTILLED
WATER
SUPPLY
SUPPLY
TUBE
DRY-BULB
THERMISTOR
AIR IN
Figure 7, Arrangement of wet-bulb, dry-bulb psychrometer.
The wet-bulb, dry-bulb psychrometer may be used up to dew
points of 100°C with reasonable accuracy (less than 5% maximum
error). Other methods of moisture measurement such as cooled-
mirror dew-point devices, hygroscopic salt devices, and semi-
conductor devices are limited by maximum ambient temperature,
usually about 50°C.
Pressure measurement—Three pressure channels are required
in the PPSS: pitot tube differential pressure (stack gas velo-
city) ; pressure at the psychrometer; and pressure at the flow
measurement section to allow pressure compensation of flowrates.
Pressure within the stack to allow calculation of total stack
volumetric or mass discharge is measured occasionally by connect-
ing one leg of the pitot tube to one of the existing transducers.
The operating ranges of these instruments are listed in Table 2.
Datametrics variable-reluctance, differential-pressure trans-
ducers with miniaturized carrier demodulators which supply dc
voltage to the calculating/display system were used to monitor
pressure. These transducers are lightweight and have displayed
good stability.
Temperature measurement—All temperatures except stack tem-
perature, probe temperature, and extension temperature are sensed
by Yellow Springs Thermilinear thermistors which show good linearity
and response. They are delicate and must be potted with epoxy
to make them more rugged and resistant to acid. The accuracy
of the measuring circuit is generally within ±0.2°C over the
range 0-93°C. Stability appears adequate, showing standard devia-
95
-------
tion of 0.05°C on the difference between two thermistors in an
oil bath. This order of stability is required by psychrometric
calculation which is based on wet-bulb, dry-bulb depression.
Sensitivity of the thermistor circuit is set at 0.1 V/°C, far
exceeding response of any thermocouple.
Stack temperature, probe temperature, and extension tempera-
ture sensors are resistance thermometers (RTD) (Yellow Springs
Platinum RTD 0-138 AX), which provide the additional range re-
quired at these locations.
Calculating/Display System
A calculating/display system is an integral part of the
PPSS.6 The system incorporates a microcomputer and microprocessor
to calculate stack volumetric flowrate, sample flowrate, iso-
kinetic ratio, and various temperature and pressure compensation
factors. A block diagram of the system is shown in Figure 8.
The system operates on 115-Vac, 60-Hz line power and retains
read-only (program) memory in the event of power loss. It dis-
plays data in large (1.3-cm) liquid crystal displays (LCD) visible
in strong sunlight. Isokinetic ratio is updated and displayed
whenever other values are not demanded. All other variables
are displayed on demand through the keyboard. Separate clocks
provide elapsed time and interval time capability (interval time
prompts move to new sampling location).
The National Semiconductor MM 57109 MOS/LSI number-oriented
microprocessor performs the number processing/calculation. This
28-pin dual in-line package provides scientific calculator in-
structions (key level language) with reverse polish notation
entry. The capabilities of this device include all of the func-
tions available on the Hewlett Packard HP-21 hand calculator.
The Intel ASM 48 single-component microcomputer performs data
sequencing computation processing and display updating.
The microcomputer is programmed in assembly language and
allows operation in the following selected modes:
a. Load constants - allows loading of all constants specific
to the run (i.e., barometric pressure, gas constants, etc.);
b. Profile - allows velocity traverse across the stack
while calculating stack velocity;
c. Static pressure - allows stack pressure to be measured
using the transducer usually assigned to monitoring psychrometer
pressure;
d. Main - starts the clocks and updating of all calculated
var iables;
e. Main pause - allows temporary interruption of the run
to relocate the sampler to another port; also accommodates mid-
run delays for any other reason.
96
-------
<§•
Po
Z.6
2
O
CO
co >
°i
>
03
O
CD
o"
1
m >
ii
O
O
m
co
CO
OTv
Om
z.<
HCD
33O
O>
I-3D
O
0>
m
3)
H
m
33
co
s
I"
MICROCOMPUTER
r
Co
00
o<
m
33
HO
-1
m
O m
m
33
-------
Although interfacing the microcomputer with all analog
signals has been completed and all calculations checked for
accuracy, overall system errors over typical ranges of operation
remain to be determined. The microcomputer has functioned with-
out failure since original debugging of its software. Its opera-
tion is rapid, with maximum calculating time 7.5 sec for updating
stack velocity.
Equipment and Structural Arrangement
The vacuum pump selected for the PPSS is a Cast 1022 carbon
vane rotary pump, which provides 280 2,/min at 0 mmHg and 68 Jl/min
at -500 mmHg. Maximum pressure drop in the PPSS is 415 mmHg
(100 mmHg occurring in the nozzle, 115 mmHg in the 200-mm filter
holder, and 200 mmHg in the dryer). The Model 1022 pump weighs
23.6 kg (52 Ib) and is the heaviest component in the PPSS. In
general, weight of any other major component is limited to about
14 kg (31 Ib).
The probe of the PPSS will be supported by an I-beam canti-
levered outward from the stack. This arrangement, shown in its
basic form in Figure 1, will be similar to Method 5 arrangement.
The probe is clamped in two places by a trolley device which
straddles the filter holder.
The sampling probe is a straight 2.32 cm ID tube of type
304 stainless steel. Total heated length is 130 cm. The probe
is heated in the manner of EPA Method 5, controlled by separate
automatic controller, and is provided with a cooling air blower
for temperature reduction to 120°C when sampling intermediate
temperature stacks. The probe extension is about the same weight
and length as the sampling probe and allows 3m total length.
Lightweight dryers containing silica gel desiccant are pro-
vided for both the main and bleed streams. The dryer is designed
for either air or ice bath cooling. Silica gel crystals are
packed within stainless steel bellows tubing. A change in mass
of the dryer can be used as a measure of total moisture in the
sample gas stream. Recharging is accomplished by replacement
of desiccant or drying in a warm oven (70CC) .
SUMMARY AND CONCLUSIONS
A prototype particulate stack sampler has been developed
with the following primary features: (1) nominal 56 Jl/min sampl-
ing flowrate, (2) electronic transducers for pressure, tempera-
ture, flow, and humidity ratio, (3) stainless steel surfaces
and components not subject to breakage, (4) electronic calcu-
lating/display system providing direct display of updated data
such as stack velocity, sample volume, and isokinetic ratio,
98
-------
(5) single-point particle size classification in the nozzle,
(6) continuous adjustability of nozzle inlet area, and (7) manage-
able weight of individual components. A wet-bulb, dry-bulb
psychrometer provides continuous indication of moisture content
of the gas stream. The psychrometer is a particularly promising
device in the severe conditions of high temperature (95 to 100°C)
and high humidity in which the device must operate.
The variable inlet, virtual impactor nozzle has potential
for well-characterized separation of fine and coarse particulate,
although versions tested to date have shown collection efficien-
cies which do not exceed 77%, and high wall losses. A new design
with symmetry in flow passages and less dimensional constraint
is recommended.
In general, the PPSS has the operating capability of the
existing Method 5 sampling train, has modernized its design,
and has eliminated the obvious problem areas. However, field
demonstration of the prototype might reveal other problems in
the novel areas of the PPSS design. The effect of acid on the
psychrometer wick and of presence of COif SOa? and other contami-
nants on behavior of the psychrometer is not known. Also, stain-
less steel surfaces in the PPSS may provide greater potential
for formation of sulfate aerosols than in the glass components
of the Method 5 train. Component ruggedness remains to be proven
by field testing.
ACKNOWLEDGEMENTS
We would like to acknowledge the assistance of the following
personnel in developing the PPSS design: Dwayne C. Ethridge,
LASL, and Professors Forney and Ravenhall of the University of
Illinois, Urbana-Champaign.
This work was supported by the Environmental Protection
Agency and performed at the Los Alamos Scientific Laboratory
operated under the auspices of the U.S. Department of Energy,
Contract No. W-7405-ENG-36.
Reference to a company or product name in this paper does
not imply approval or recommendation of the product by the Uni-
versity of California or the U.S. Department of Energy to the
exclusion of others that may be suitable.
99
-------
REFERENCES
1. Federal Register 36(247)(December 23, 1971).
2. Elder, J.C., M.I. Tillery, and H.J. Ettinger. Evaluation
of EPA Method 5 Probe Deposition and Filter Media Efficiency.
Los Alamos Scientific Laboratory Report LA-6899 PR, August,
1977.
3. Forney, L.J. Aerosol Fractionator for Large-Scale Sampling.
Rev. Sci. Instrum. 47 (10):1264-1269, October 1976.
4. Forney, L.J., D.G. Ravenhall, and D«S. Winn. Aerosol Im-
pactors: A Study of a Fluid Jet Impinging on a Void. J.
Appl. Phys. 49(4):2339-2345, April 1978.
5. Federal Register 41(187) (September 24, 1976).
6. Ethridge, C.D., and L.G. Littlefield. Microcontroller for
Exhaust-Stack Environmental Measurements. Los Alamos Scien-
tific Laboratory Report LA-UR-79-968, June 1979.
100
-------
PAPER 6
THE EFFECTS OF NOZZLE LOSSES ON IMPACTOR SAMPLING
KENNETH T. KNAPP
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
U.S. ENVIRONMENTAL PROTECTION AGENCY-RTP
When measuring the mass, size distribution, and other prop-
erties of aerosols, getting a representative sample to the mea-
suring device system is an absolute must for obtaining good data.
The retention of particles by any part of the sampling system
can greatly bias the results. With impactors and other particle
sizing devices, the retention by the nozzle may be particle de-
pendent and, therefore, completely change the size distribution
obtained. This problem has been addressed by Marple and Willeke1
and others2"1* but few data have been given.
The retention in the nozzle and the probe in the standard
stationary source standard particulate emission measurement
methods such as EPA Methods 5 and 17 does not cause a problem
since the nozzle and probe catches are included in the sample.
However, this is not the case with the proposed inhalable par-
ticulate matter (IP) measurement method. The IP standard is
to include all particulate material with an aerodynamic size
at ambient conditions of 15 ym or less. The stationary source
IP emission measurement method must contain a 15 ym particle
size cut system. The general approach is to use a 15 ym cut
point preseparator on the nozzle end of the probes used in the
particulate mass measurement methods such as EPA Method 5. The
preseparator would remove the particles larger than 15 ym aerody-
namic size and allow the remaining particles to pass into the
collection system.
Experiments were designed to test such a separation system.
The first set-up tested was the preseparator with a button-hook
nozzle, the nozzle type most often used with EPA Method 5 and
other types of source sampling trains, since these nozzles will
fit into the 3 or 4-inch ports most commonly available. A back-
up filter was used for the collection of the various size and
type aerosols. The efficiency of the preseparators was deter-
mined by comparing the amount of aerosol retained in the pre-
separator including the nozzle to the total amount of aerosol
101
-------
collected. After overcoming problems in generating monodisperse
aerosols with aerodynamic diameters greater than 15 yin,. testing
began. On the first test runs with monodisperse aerosols of
about 10 urn, the button-hook nozzles were found to collect more
material than the preseparators. If these nozzles retained a
large amount of material in the size range less than 15 ym, then
they cannot oe used with the preseparators without introducing
serious errors in the IP measurement. Before a decision on
whether the button-hook nozzles could be used with the IP method,
the extent of the retention of the aerosols smaller than 15 ym
by these nozzles needed to be studied. Such a study was under-
taken with 3/16, 1/4, and 5/16-inch button-hook nozzles. These
were chosen for the tests since they are the size nozzles most
often used in stationary source testing. In addition, a 90°
bend 1/4-inch nozzle was used in some tests.
Tests with laboratory generated monodisperse aerosols and
re-dispersed sized coal-fired power plant fly ash were made.
The monodisperse aerosols were methylene blue particles generated
by a vibrating orifice (Berglund-Liu) aerosol generator. The
aerodynamic size range covered with these aerosols was from about
3 ym to about 27 ym. The coal-fired fly ash used was separated
into five narrow size range fractions which had mass median diam-
eters (MMD) of 3.5, 5.8, 8, 16.5, and 26.7 ym. The sized fly
ash fractions were re-dispersed by the aerosol generator of the
Stationary Source Simulator Facility (SSSF) in which the nozzles
were tested.
The results from several test runs with the %-inch button-
hook nozzle and the monodisperse aerosols are given in Table 1.
TABLE 1. RETENTION OF MONODISPERSE AEROSOLS
BY 1/4-INCH BUTTON-HOOK NOZZLE
Aerosol diameter, ym % in %-inch nozzle
3.
5.
9.
10.
11.
15.
16.
26.
3
5
3
5
6
1
3
7
27
53
79
901:
13*
8*
6
4
+
+
+
D
3
D
+
+
2a
3a
6a
ia
la
a
Average and range of 3 runs
Single runs
102
-------
The results of runs with the three nozzle sizes on various
aerosols are given in Table 2.
TABLE 2. COMPARISON OF RETENTION OF
MONODISPERSE AEROSOLS BY THREE NOZZLE SIZES
Aerosol diameter, ym
6.4
9.3
10.5
11.6
12.8
13.9
15.1
18.6
21
24
3/16
21
53
22
2
1
4
5
11
6
2
1/4
48
79
90
12.7
12
22
8
3
4
3
5/16
0
46
54
59
75
68
47
51
45
22
Figure 1 is a plot of the data given in Table 2.
Q3/16 INCH
A 1/4 INCH
O 5/16 INCH
2 4 6 8 10 12 14 16 18 20 22 24
AEROSOL MASS MEDIAN DIAMETER, urn
Figure 1. Retention of mondisperse aerosols by nozzles of three sizes.
103
-------
A comparison of the ^s-inch button-hook with a 90° bend \-
inch nozzle was made with monodisperse aerosols of 13.9 jam diam-
eter. The results were 22% retained in the button-hook and 24%
in the 90° bend nozzle thus indicating no difference.
The results from the fly ash study are given in Table 3
and Figure 2.
TABLE 3. RETENTION OF FLY ASH BY NOZZLES
OF THREE SIZES
Aerosol MMD, ym
3.5
5.8
8
16.5
26.7
3/16
28
22
6
4
2
% in nozzle
1/4
27
42
33
6
4
5/16
16
14
41
29
26
The tests were for collection periods of 10 minutes. The
question whether the nozzle collection would reach an equilibrium
was considered and runs of different collection times were made
with the ^-inch button-hook nozzles. The results of these tests
are shown in Table 4.
100
90
80
s? 70
z
ji eo
ui
m 50
oc
j 40
N
N
i 30
20
10
0
T
A 3/16 INCH
O 1/4 INCH
O 5/16 INCH
I
4 6 8 10 12 14 16 18 20 22 24
FLY ASH FRACTION MASS MEDIAN DIAMETER, jum
26
Figure 2. Retention of fly ash by nozzles of three sizes.
104
-------
TABLE 4. EFFECT OF COLLECTION TIME ON
RETENTION OF FLY ASH BY I/4-INCH NOZZLE
Collection time, min % in nozzle
5 44
10 33
20 48
30 32
No major difference over the period tested was found.
All of the above data are from runs where the flow rate
was about 0.014 m3/rnin (Q.5 CFM) . This is the flow rate at which
the preseparators and impactors operate most efficiently. How-
ever, since some tests will be made at different flow rates,
a study of the effects of flow rate on nozzle retention was car-
ried out with the %-inch button-hook nozzles. Three flow rates
were chosen for the test: the optimum flow rate for the pre-
separators and impactors being tested (0.5 CFM); the upper limit
of efficient operation of the preseparators (0.75 CFM); and a
flow rate commonly used with EPA Method 5 (1 CFM). The 1 CFM
flow rate is not recommended for the preseparators because at
this flow rate severe re-entrainment occurs resulting in the
preseparator passing large amounts of particles greater than
15 urn. The results for tests at the three flow rates with an
aerosol of about 11 ym are given in Table 5.
TABLE 5. FLOW RATES VS RETENTION OF FLY
ASH BY 1/4-INCH NOZZLE
Flow rate, m3/min % in nozzle
0.014 (0.5 CFM) 20
0.021 (0.75 CFM) 4
0.028 (0.99 CFM) 1
The results of these tests have shown that the button-hook
and 90° bend nozzles retain various amounts of the aerosols that
are drawn through them. The amount retained depends on the
nozzle size, the aerosol size, and the flow rate. The percent
collected in the nozzle remained fairly constant over a time
period up to 30 minutes and, therefore, the collection does not
appear to reach an equilibrium. Thus, the possible use of pre-
conditioning is eliminated.
105
-------
Because much of the aerosols retained in the nozzles should
be passed on through the preseparator and onto the filter or
other collection devices, the use of the button-hook or 90° nozzle
with the IP method will cause large and varied errors. The
nozzle retention does not cause a problem with the standard sta-
tionary source standard particulate emission measurement methods,
such as EPA Methods 5 and 17, since the nozzle catch is included
in the total catch. However, particle size measurements with
these types of nozzles will have serious errors.
ACKNOWLEDGEMENTS
The author wishes to acknowledge the assistance given by
Donald Duke and Raymond Steward for the aerosol generation studies
and Thomas Ward for separating the fly ash.
REFERENCES
1. Marple, V.A., and K. Willeke. Impactor Design. Atmos.
Environ. 10:891-896, 1976.
2. Lundgren. D., and S. Calvert. Aerosol Sampling with a Side
Port Probe. Am. Ind. Hygiene Assoc. J. 28:208-215, 1967.
3. Cheng, Y., and C. Wang. Inertial Deposition of Particles
in a Bend. J. Aerosol Sci. 6:139-145, 1975.
4. Crane, R.I., and R.L. Evans. Inertial Deposition of Par-
ticles in a Bent Pipe. J. Aerosol Sci. 8:161-170, 1977.
106
-------
PAPER 7
DILUTION SOURCE SAMPLING SYSTEM
ROBERT J. HEINSOHN
JOHN W. DAVIS
CENTER FOR AIR ENVIRONMENT STUDIES
THE PENNSYLVANIA STATE UNIVERSITY
AND
KENNETH T. KNAPP
NATIONAL ENVIRONMENTAL RESEARCH LABORATORY-RTF
U.S. ENVIRONMENTAL PROTECTION AGENCY
ABSTRACT
A source sampling system has been designed and tested that
is lighter and easier to use than the conventional EPA Method 5
sampling systems. The heart of the system is an ejector pump
that uses dry air to simultaneously pump, cool, and dilute a
sample of a process gas stream. The sample is treated as a minis-
cule plume and the particles are removed from the sample after
it has been cooled and diluted with dry air.
Simultaneous, full-scale source certification tests were
run with the dilution sampling system and a conventional EPA
Method 5 system. The emission from a coal-fired steam boiler,
a glass melt tank equipped with an electrostatic precipitator,
and a lime kiln equipped with a fabric filter were tested. Test
results reveal that:
(a) the dilution system was considerably easier to use;
(b) the particle mass concentrations obtained with the
dilution system were larger than those obtained with
an EPA Method 5 system;
(c) the size distribution obtained with the dilution sys-
tem contained submicron particles not observed at stack
conditions as measured with an in-stack impactor;
(d) the relative abundance of elements in particles ob-
tained by the dilution sampling system was considerably
different from that in comparably sized particles
obtained by the EPA Method 5 system. No consistent
trends could be determined.
107
-------
NOMENCLATURE
a,b constants
c apparent concentration of a material per unit
m'° mass of particle at any time zero
c concentration of a material per unit surface
s area of particle
c , c mass concentration of a material per unit volume
v v'e of the carrier gas at any time t, or at such
time when no further transport occurs, i.e.,
its equilibrium value
D diffusion coefficient
D particle diameter
g acceleration of gravity
hf enthalpy of vaporization
H effective stack height
m^, m mass flow rate of dilution air, sample
M , M molecular weight of air and sample
cl S
n number of particles per unit volume of carrier
gas
Pr, Pr' Prandtl number
P., P , PQ, psat pressure of supply air, diluted sample, control
1 pressure, vapor saturation pressure
AP pressure difference across the sample orifice
S
Q emission source strength
Q , Q volumetric flow rate of dilution air and gas
s sample
R gas constant
r dilution ratio (Qd/Qs), each corrected to
standard conditions
Re Reynolds number
s atmospheric stability parameter
108
-------
S degree of saturation
T,, T , T temperature of dilution air, diluted sample,
d m s sample
T, boiling temperature of a liquid
t time
u wind speed
X mole fraction
x,y,z spatial coordinates
a,B constants
6 potential temperature
K thermal conductivity
p , p0 density of vapor and of the liquid
v &
a , a atmospheric dispersion coefficients
Y z
T time constant
INTRODUCTION
An inherent limitation in all contemporary source sampling
systems is their inability to obtain particles in the state they
will have when the plume mixes with the atmosphere. In-stack
devices capture particles at stack gas temperature and gas com-
position and EPA Method 5 systems capture particles on filters
at 121°C and in impingers at 0°C. In these cases condensation,
adsorption, and agglomeration that occur in the plume do not
occur in a similar fashion in the sample. The need for a sampl-
ing system that will enable operators to capture particles that
represent those a plume transfers to the atmosphere has long
been recognized and is the basis for a novel source sampling
system reported in this paper.
A source sampling system with dilution (hereafter abbreviated
SSD) has been designed that treats the sampled gas as a miniscule
plume and removes particles after the sample has been diluted
with clean dry air.1"1* The purposes of this paper are to:
(1) describe the design and operation of the SSD system;
(2) analyze simultaneous samples taken with the SSD system
and a conventional EPA Method 5 system; and
(3) compare the processes of adsorption, condensation,
and agglomeration in the SSD system and plume.
109
-------
DESIGN OF SSD SYSTEM
The SSD system is shown schematically in Figure 1; it con-
sists of six components.
1. Probe; (length 2.07 m; mass 4.14 kg) null reading iso-
kinetic nozzle, impactor preseparator, heated probe, 1.27 cm
(^-inch) inside diameter, 1.83 m (6 ft) long.
2. Pump and Filter Assembly; (length 0.68 m; mass 7.26
kg) ball valve, sample orifice, heated enclosure, Dilution Ejector
Pump (DEP), filter holder.
3. Umbilical Cord; dilution air line, 120 VAC control
cable, null nozzle pressure lines, sample orifice pressure lines,
thermocouple leads.
4. Control Unit; (0.46 m x 0.33 m x 0.23 m; mass 20.3
kg) pressure transducer, pressure regulator, pressure gauges,
temperature controller, critical flow orifice, dilution air
filter.
5. Flow Measurement Unit; (0.38 m x 0.30 m x 0.23 m; mass
7.3 kg) electronics for indicating elapsed time, volumetric flow
rate, and volume of sample.
6. Dilution Air Cleaner and Dryer; (0.79 m, 0.41 m, 0.40
m; mass 25.7 kg) filters to remove oil and foreign matter, re-
generative dryer to remove water vapor.
Units 1 and 2 are fitted with a taper fitting and mounted
on a tripod for insertion into the stack. The Control Unit
houses all the necessary meters and valves for controlling the
flow of dilution air. The Flow Measurement Unit consists of
power supplies and digital logic system for conditioning, inte-
grating, and storing the signals related to the sample volumetric
flow rate and elapsed time. The Dilution Air Cleaner and Dryer
cleans and dries air to operate the system. The unit is located
on the ground, and the air is supplied to it by plant air service
or a portable air compressor. The unit is composed of commercially
available equipment and will not be discussed.
Performance specifications for the SSD system are as follows:
I) Sample flow rate: 0.236 to 0.708 liter/s (0,5 to 1.5
sofm) and is adjustable so that isokinetic sampling conditions
can be achieved at the nozzle inlet.
2) Sampling nozzles: 0.318 cm, 0.635 cm, 0.953 cm, and
1.270 cm (1/8 in., 1/4 in., 3/8 in., 1/2 in.).
3) The gas sample and dilution air are completely mixed
before the particles are removed by a filter; the temperature
of the mixture is within 5°C of the atmospheric temperature.
4) The entire apparatus is easily cleaned.
110
-------
IMPACTOR
PRESEPARATOR
PROBE
ASSEMBLY
TAPER FITTING
PUMP&FILTER
ASSEMBLY
BALL VALVE
HEATED CANISTER
/
COMPRESSED
AIR IN,
HEATED SAMPLE
PROBE
NULL-READING ISOKINETIC
SAMPLING NOZZLE
I
I DEP
I
UMBILICAL
SAMPLE
FLOW MEASUREMENT,
UNIT
FLOW, VOLUME |
SAMPLE TIME T
READOUT I
r
i
AIR CLEANER AND DRYER
I I
~~!
1 COARSE
I FILTER
I
SE FINE
R FILTER
1
K
K\
'
1
1
1
i
..„>..
DRYER 1
1
'XA
|
ORIFICE
TRANSDUCER
THREE-WAY VALVE
=e—
Ar
PRESSURE
| REGULATOR
I
si
Pf. T,
.PRESSURE
/REGULATOR
47 mm FILTER
/TEMPERATURE
CONTROLLER
-CONTROL
UNIT
-FILTER
Figure 1. Schematic diagram of sampling system.
-------
Dilution Ejector Pump—The Dilution Ejector Pump (DEP) shown
in Figure 2 is the heart of the system. Sample gas is pumped
by a high velocity stream of dry dilution air passing through
the annular space around the sample tube, transferring momentum
to the slower moving sample gas stream. The two streams mix
and achieve uniform conditions5"8 at the end of the mixing tube.
The diluted stream then passes through a filter to remove par-
ticles. The DEP was designed to use a minimum amount of air
to cool the sample yet keep it above the dew point. Figure 3
shows the "well mixed" sample temperature as a function of the
dilution ratio (on the basis of volume), inlet sample temperature,
and moisture content. The SSD system was designed to accommodate
dilution ratios of 10 to 15.
Control of sample flow rate—Dilution air enters the Control
Unit, is regulated to a set pressure (Po), passes through a three-
way ball valve and an orifice and into the DEP. The dilution
air filter assures that all the particles collected on the DEP
filter are from the sample gas. The pressure, Po, controls the
sample flow rate and is controlled by the pressure regulator.
The supply pressure is P^. A three-way ball valve is used to
pass the dilution air to the DEP (as shown) or to a separate
calibration system. The sample mass flow rate (ms) is controlled
by the mass flow rate of the dilution air, m^. Since the geo-
metry of the DEP is constant, the mass flow rate of the dilution
air is a function of the regulated pressure, PQ. The pressure
of the mixed gas before the filter (P^) affects the sample mass
flow rate, but as this pressure changes only slightly as par-
ticles are collected on the filter, the pressure, Po, is the
principal factor controlling the sample flow rate.
1. MIXING AND DILUTION CHAMBER
2. NIPPLE
3. SAMPLING TUBE
4. SAMPLING TUBE POSITIONER
5. MOUNTING PLATE
6. TUBE TO MALE PIPE FITTING
7. GASKET
Figure 2. Scale drawing of the dilution ejector pump (DEP).
112
-------
150
o
o
120
100
00
Q
LU
LL
O
UJ
cc
D
cc
UJ
Q.
5
UJ
WATER VAPOR IN SAMPLE
18.9% BY VOLUME
10 12 14 16
18
20
DILUTION RATIO, r
Figure 3. Temperature and relative humidity of diluted sample vs. dilution
ratio for dilution air at 27°C and relative humidity less than 10%.
The sample flow rate also depends on the size of the nozzle
and the length of the probe. Figure 4 is a graph of sample volu-
metric flow rate (Qs) and the dilution ratio (Qf3/Qs) versus the
control pressure PQ. For the BSD system, accurate measurement
of dilution volumetric flow rate is not needed since the operator
adjusts it to achieve isokinetic sampling conditions and a diluted
sample temperature within 5°C of atmospheric temperature. It
is essential that the sample volumetric flow rate be measured
accurately.
Measurement of volumetric flow rate and total volume of
sample—An electronic system is used for measuring the total
volume of the sample and the sample volumetric flow rate, and
in displaying these values. An orifice (0.635 cm diameter)
mounted upstream of the DEP was used to measure the sample volu-
metric flow rate. The temperatures of the orifice housing and
probe were held at 93.3°C (200°F). The electronic signal is
generated by a pressure transducer (located in Control Unit)
that senses the difference in pressure across the orifice (APS).
The transducer output is a 0.5 to 5.5 volt signal that is linearly
proportional to the differential pressure of 0.0 to 38.1 cm H20
(0.0 to 15 inches H20).
113
-------
o
o
CM
O)
1 «=-
E b
§s
Si
£•?
c « 5.0
S E
4.0
<
CO
3.0
2.0
I
I
O 0.64 cm
D 0.95 cm
A 1.27 cm
I
I
I
I
I
I
I
I
I
17
16
15
14
13
•
a
(0
11
10
9
{£
Z
O
10 11
12
123456789
CONTROL PRESSURE Po(N/m2) x 104
Figure 4. Operating characteristics of the DEP using a glass fiber filter
and different inlet nozzles.
The volumetric flow rate of the sample, Qs, is proportional
to the square root of AFS, which is obtained by passing the trans-
ducer output through a square root extractor. The output volt-
age from the square root extractor indicates the volumetric flow
rate that is needed in the computation of sampling velocity at
the nozzle. Integrated with time this signal determines the
total gas volume of the sample. The integration is achieved
by converting the output voltage from the square root extractor
to a series of pulses with an analog-to-frequency converter,
and then summing these pulses with respect to time. The total
number of pulses is counted, and displayed as the total volume
of the sampled gas. In addition, the elapsed time is also dis-
played. The display is a set of light emitting diodes (LED's).
Figure 5 is a schematic diagram of the entire electronic system.
114
-------
Ul
COUNTING BOARD (VOLUME)
RESET
TRANSDUCER
I 1 1 i I i
FREQUENCY
GENERATOR
PROBE INPUT BOARD
COUNTING BOARD (TIMES)
MULTIPLEXER BOARD
Figure 5. Block diagram electronic circuit.
-------
Four counting channels are used, two for counting elapsed
time, and two for counting the sampled volume of gas. Each
volume counting channel is associated with one of the time count-
ing channels. The information stored in each set (one volume/one
time channel) can be initialized to zero by a button. In the
field, one volume and time set is initialized to zero before
the test, while the other is initialized to zero after sampling
at each traverse point. One LED display indicates volume, and
one LED display indicates elapsed time. A selector switch deter-
mines which set of information is displayed: volume and time
for entire test or volume and time for that traverse point.
Calibration and cleaning system—By means of two ball valves,
the operator can withdraw a sample from the stack or can pass
clean dry air through the system at a desired rate so as to re-
move particles from the orifice and DEP or to check the calibra-
tion of the orifice. The dislodged particles pass through the
system and are deposited on the filter. Thus no particles are
lost during the cleaning and calibration operation which can
be performed without removing the probe from the stack. The
numerical value of the sampled gas volume is electronically
stored prior to cleaning and calibration and can be recalled
once sampling has resumed. If particles collect around the
orifice and change its discharge coefficient, it will be detected
by the calibration procedure. To compensate for a change in
calibration, a "calibration adjust" potentiometer mounted on
the face of the flow measurement unit enables the operator to
adjust the recorded flow rate and to set it equal to the cali-
bration flow rate.
OPERATION IN THE FIELD
In the field the sampling system is operated as follows:
1. A nozzle is selected such that the sample flow rate
will be between 0.0140 and 0.0425 normal m3/min. (Q.5 to 1.5
scfm). (If a null-reading nozzle is not used, an independent
S-type pitot must be attached to the sampling probe.)
2. With the ball valve closed, the probe is inserted in
the stack. The stack gas velocity then is determined. The three-
way valve is set and dilution pressure adjusted to pass dilution
air through the DEP at a rate that will withdraw a sample at
the isokinetic sampling rate.
3. The ball valve is then opened allowing the sampled gas
to flow through the orifice. The dilution air flow rate is read-
justed to maintain isokinetic sampling.
116
-------
COMPARISON TESTS
Several samples were taken simultaneously with the SSD
system and an EPA Method 5 sampling train from the following
sources:
1) Coal-fired, traveling grate stoker, boiler (tests 1-5);
2) Glass melting tank (tests 6-11);
3) Lime kiln (tests 12-14).
In tests 1, 6, 7, and 12-14, simultaneous full-scale source
certification tests were conducted. Tests 2-5 and 8-11 were
conducted to determine the size distribution and elemental analy-
sis of the particles. In tests 2-5 particle size measurements
were made with an Andersen-2000, 8-stage, in-stack impactor and
the SSD system. In tests 8-11 the size distributions were deter-
mined by scanning electron microscopy. Table 1 shows the mass
concentrations that were obtained, Table 2 shows the size distri-
bution, and Table 3 shows the results of elemental analysis per-
formed by a scanning electron microscope and an X-ray spectro-
graph. Since the scanning electron microscope has poor accuracy,
the ratio of the amount of the element in the SSD sample to that
in the in-stack sample was computed.
TABLE 1. COMPARISON OF THE MASS CONCENTRATION DETERMINED BY
THE SSD SYSTEM WITH THAT DETERMINED BY METHOD 5 SYSTEM
Source
Test
Loading (mg/ncm)
SSD
EPA Method 5
with impingers
EPA Method 5
without impingers
Coal
fired
boiler
Glass
melt
tank
Lime
kiln
6
7
12
13
14
495.6
71.7
351.7
116.2
99.4
139.0
462.1
45.8
331.8
126.6
83.8
140.3
407.6
38.9
323.6
111.0
70.0
123.6
117
-------
TABLE 2. MEAN DIAMETERS AND GEOMETRIC STANDARD DEVIATIONS
OF PARTICLES OBTAINED WITH THE SSD SYSTEM AND
AN IN-STACK IMPACTOR
Source Test Sampling train
,•
2
In-stack
SSD
Coal
fired • 3
In-stack
SSD
boiler
4
5
-
Glass
melt 8
tank
In-stack
SSD
In-stack
SSD
In-stack
SSD
Dry gas
volume
(ncm)
0.335
1.902
0.262
1.376
0.295
1.351
0.312
1.389
0.192
1.000
Mean diameter
(Vim)
7.2
0.4
10.2
3.0
9.6
0.3
4.1
0.4
0.4
0.4
Geometric
standard
deviation
3.1
6.6
2.4
5.4
3.9
1.7
4.6
5.3
2.2
2.2
DISCUSSION OF RESULTS
Apparatus—One of the primary assets of the SSD system is
its size and ease of operation. The DEP is considerably lighter
and smaller than the mechanical pump in an EPA Method 5 system.
The SSD has no glassware, impingers, ice, or impinger box to
be carried to the sampling site. With no impinger box, the SSD
system is also easier to install on a support mechanism, such
as a monorail. The dilution air cleaner and dryer can be left
unattended on the ground. The ball valve located after the probe
and before the sample flow orifice is a positive shut-off. When
a large probe is used the ball valve may be beyond the catwalk
or roof and require the operator to reach over the railing.
If the valve cannot be reached, the system can be shut off by
bringing the dilution air pressure to zero. This, however, is
not a positive shut-off.
Deposition of particles in the sampling system—Deposition
of particles in the sampling system (probe, orifice and DEP)
proved to be less than what occurs in an EPA Method 5 system.
A "scalping cyclone" (Andersen-2000 preseparator) was mounted
between the sampling nozzle and upstream end of the heated probe
and removed particles greater than approximately 10 pm. Table 4
shows the percentage deposition in parts of the SSD system for
a number of tests.
118
-------
TABLE 3. RELATIVE ABUNDANCE OF ELEMENTS IN THE PARTICLES
OBTAINED BY THE SSD SYSTEM TO THAT ON THE PARTICLES
WITHIN THE STACK
Test No.
Impactor
stage
3
4
5
6
7
8
Back Up
, 2 Coal-Fired Boiler with Aluminum as Reference Element
Cut size
ym
SSD/In-
stack
6.95/6.65
4.80/4.50
3.10/3.00
2.05/1.98
1.05/0.96
0.61/0.60
0.42/0.40
Element
Na
Al Si S
1.09 1.
0.39 1.
0.15 1.
2.18 1.
1.65 1.
4.76 1.
1.44 1.
Test No.
,0 1.14 1.
,0 0.63 0.
,0 0.69 2.
,0 1.47 1.
,0 0.96 0.
,0 0.80 0.
,0 0.95 0.
34
96
02
46
35
30
24
9 Glass Melt
K
1.02
0.80
1.62
1.02
0.45
1.79
0.90
Tank
Ca Fe Zn
1.
0.
0.
1.
1.
0.
2.
59 0.88
49 0.69
98 2.41
57 0.77
59 0.40 0.0
50 0.08 0.0
00 0.67 4.51
Element
Dp (ym)
6.30
4.00
2.50
1.60
1.00
0.63
Background
1
1
1
1
1
1
.00
.00
.00
.00
.00
.00
Al
0
0
0
0
0
1
Test
.77
.88
.97
.89
.99
.00
No.
Si
0.62
0.94
1.01
1.00
1.02
0.92
S
1.38
1.16
0.96
0.88
0.66
K Ca
0
0
0
0
0
0
10 Glass Melt
.58 0
.83 0
.88 0
.91 0
.83 0
.87 0
Tank
.67
.92
.97
.84
.95
.74
Fe
0.92
0.93
1.03
0.97
1.02
0.99
Ti
0.72
0.99
1.03
1.08
0.94
0.81
Element
Dp (ym)
10.00
6.30
4.00
2.50
1.60
1.00
0.63
Background
1
1
1
1
1
1
1
.00
.00
.00
.00
.00
.00
.00
0
0
1
1
1
1
1
Test
Al
.50
.73
.06
.18
.00
.10
.09
No.
Si
1.71
0.18
1.22
1.16
1.02
1.00
1.10
S
1.50
2.70
0.59
0.50
0.68
1.02
1.30
0
0
1
1
0
1
0
11 Glass Melt
K
.71 0
.82 1
.10 1
.34 1
.98 0
.09 1
.95 0
Tank
Ca
.79
.14
.06
.38
.96
.07
.93
Fe
0.78
1.01
1.48
1.50
1.10
0.94
1.15
Ti
0.75
0.96
1.14
1.27
1.07
1.00
1.16
Element
Dp (ym)
10.00
6.30
4.00
2.50
1.60
1.00
0.63
0.40
Background
1
1
1
1
_i_
1
J_
1
1
.00
.00
.00
.00
.00
.00
.00
1
0
1
1
0
0
0
Al
.02
.88
.26
.27
.79
.50
.64
Si
1.18
0.82
1.60
0.65
1.17
0.47
0.83
S
1.49
0.46
1.50
0.73
1.04
0.37
0.32
0
0
0
0
0
0
0
K
.60 0
.57 0
.91 1
.88 1
.80 2
.51 1
.78 6
Ca
.20
.80
.14
.86
.29
.77
.53
Fe
1.02
0.82
1.35
0.35
0.14
0.24
0.22
Ti
0.96
1.51
4.20
0.87
1.02
0.56
1.03
119
-------
TABLE 4. DEPOSITION IN SSD SYSTEM (IN PERCENT)
Nozzle
Filter and DEP and preseparator
Test No. Source filter housing orifice and probe
1
6
7
12
13
14
Power plant
Glass melt tank
Glass melt tank
Lime kiln
Lime kiln
Lime kiln
35.1
16.7
79.0
36.6
48.0
36.5
18.2
41.5
7.0
18.7
21.5
16.5
46.7
41.8
14.0
44.7
30.5
47.0
In laboratory experiments a monodisperse aerosol of uranine
dye was drawn through the sampling system at a volumetric flow
rate of 28.32 normal liters per minute (1 scfm). It was shown .
that
(a) the scalping cyclone has a cut size of approximately
5 ym and is capable of removing particles above 10 pm,
(b) deposition in the sampling nozzle and orifice meter
is never more than 10% of the sampled aerosol,
(c) deposition in the DEP is never more than 20% of the
sampled aerosol,
(d) the final filter is the principal agent removing par-
ticles that leave the scalping cyclone.
Water content of samples—One important difference between
the SSD system and an EPA Method 5 system is the procedure to
calculate the water content of the sample gas. EPA Method 5
trains condense the water vapor and measure the remaining volume
of gas in a positive displacement meter. The water content of
the sample gas is easily determined by measuring the water in
the impingers. The SSD system measures the total volume of
sampled stack gas including the water vapor. The SSD system
thus requires a separate procedure (such as EPA Method 4) to
determine the water content.
Isokineticity is inherently easier to achieve in the SSD
system. An EPA Method 5 train uses a pitot tube to measure the
stack gas velocity and the sample flow rate is adjusted to achieve
isokinetic conditions. To calculate the sample volumetric flow
rate to achieve isokinetic conditions, the temperature and water
120
-------
content of the gas entering the nozzle must be known. Since the
sample gas flow rate is measured after the water vapor has been
condensed, the water content must be assumed to calculate the
velocity. The SSD system measures the actual volumetric flow
rate before condensation of water vapor and the above assumption
does not have to be made.
Particle size distribution, mass concentration and elemental
analysis—The mass concentration obtained by the SSD system was
on the average larger than that obtained by the EPA Method 5
system when the impinger catch was included and even larger than
when the impinger catch was not included. From the coal-fired
boiler exhaust, the size distribution obtained by the SSD system
contained a peak at 0.5 ym that was not present in the sample
obtained at stack gas temperature. Figure 6 is typical of the
data taken for tests 2-5. The results of the elemental analysis
show no consistent pattern.
160 —
140 —
120 —
100 —
c.
Q
01
TJ
•D
— DILUTION SAMPLING SYSTEM
IN-STACK IMPACTOR
0.5 1.0 5.0
PARTICLE DIAMETER Dp((im)
Figure 6. Particle size distribution, Test No. 3.
10.0
20.0
121
-------
It is believed that the mass concentration, size distribu-
tion, and elemental analysis are affected by condensation, adsorp-
tion, and agglomeration that occur when the process gas stream
mixes with air. The processes of adsorption, condensation, and
agglomeration were not modeled in this research so that predic-
tions of the mass concentration, size distribution, and the par-
ticle composition cannot be made. Rather, these processes will
be examined in the plume and for samples obtained with the SSD
system and an EPA Method 5 system. The purpose of this examina-
tion is to show that while the behavior of the diluted sample
and the plume differ in a number of respects, samples obtained
with the SSD system are more representative of the plume than
samples obtained by either EPA Method 5 or Method 17 systems.
A plume rises because of its momentum and buoyancy. In
plumes from combustion processes buoyancy is dominant. The plume
rises until its density equals that of the surrounding atmosphere.
Since plumes from fossil fuel burners have molecular weights
that are approximately the same as air, the plume rises until
its temperature is essentially equal to the surrounding air,
whereupon the wind then carries the plume downwind where it
expands laterally and vertically by turbulent diffusion. The
concentration of species within the plume can be expressed by
the Gaussian plume equation,
cv =
_ Q_
2irua a
Y
exp - (
exp
r
exp [ -
(Z+H)
—r-
(1)
The critical parameters in the expression are the diffusion co-
efficients tfy,az , which increase with downwind distance x and
the atmospheric instability. From the above it can be seen that:
(a) at any downwind distance the species concentration
is Gaussian with respect to the vertical (z) and trans-
verse (y) distances
(b) with respect to downwind distance the concentration
decreases exponentially.
Particles in the plume that are transferred to the atmosphere
need to be characterized in terms of mass concentration, size
distribution, and chemical composition.
The way the SSD sample mixes with dilution air and the way
the plume mixes with the atmosphere differ in at least three
respects: (a) Reynolds number, (b) time scale, and (c) mixing
mechanism.
(a) Reynolds number. The Reynolds numbers of plumes depend
on the installation but values of 10,000 or larger can be ex-
pected. The Reynolds number of the sample as it enters the DEP
122
-------
is 1000 or less. Thus not only are the Reynolds numbers dif-
ferent, but the plume is turbulent while the sample is laminar.
(b) Time scale. The rates at which the sample and plume
mix with air are also different. The sample mixes with dilution
air with a time scale given by:
time = DEP volume/volumetric flow rate * 0.5 sec.
The plume mixes with the atmosphere over a time period that is
difficult to estimate. Bosanquet9 suggested that the time for
a buoyant plume to rise is 2.16 s"1 where s is related to the
potential temperature gradient iLi by
3z
,g. 8 6 , _.
s = ~z (2)
For a weakly stable atmosphere, t is approximately 200 seconds.
Once the plume ceases to rise it spreads and mixes slowly with
a time scale of the order of hours or days. Assuming a plume
enters the atmosphere at 300°C and that the atmospheric tempera-
ture is 25°C, the cooling rate of the plume and SSD sample are,
plume: (300 - 25)/200 = 1.83°C/sec
SSD sample: (93.5 - 25J/0.5 = 137°C/sec
Thus the cooling rate for a plume is considerably slower than
the cooling rate for the SSD sample. The mixing of the plume
with the atmosphere is characterized by large scale turbulent
eddies and a slow rate of change toward equilibrium, while for
the SSD sample mixing is characterized by small scale turbulent
eddies and rapid changes toward equilibrium.
(c) Mixing mechanism. After the plume rises, it is trans-
ported downwind. All during this time it expands and mixes with
the atmosphere. Momentum transfer with the atmosphere is small
and the plume spreads by lateral diffusion. The concentration
of species within the plume can be expressed by Equation 1.
The sample on the other hand mixes with dilution air all the
while the two streams are confined within the DEP. The sample
and the coaxial stream of air mix by the transfer of momentum.
The concentration of species within the two streams can be esti-
mated by various expressions obtained from the literature of
ducted turbulent jets.5"7 The thermodynamic path followed by
the species within the plume is different from the path followed
within the sample. The SSD sample quickly mixes with dilution
air and is nearly uniform in temperature and concentration as
it enters the filter. The plume remains distinct for a consider
able time and mixing is gradual.
123
-------
The significance of these differences must be judged by
examining how particles in a process gas stream change when the
stream is cooled and diluted with air. Particles within the
plume or sample change by any one or combination of three pro-
cesses: (a) particles may combine by coagulation, or their
surfaces may act as sites for (b) condensation or (c) adsorption.
To study how well a sample obtained with an SSD system simulates
conditions within the plume it is necessary to determine if these
three processes are affected by the differences above.
Coagulation. The rate at which the concentration (n) of
particles of a particular size decreases by coagulation is given
below. (Similar expressions should also be written for smaller
particles that combine to form a new particle of the particular
size in question. The net change in concentration will then be
the difference between the two rates.) For a first estimate,
simply consider the rate of removal of particles of a certain
size ,
^ = -kn2 (3)
dt
The coefficient k depends on the size of the particles but it
is of the order of 10~10 cn^sec"1. Because k is small, and n
is of the order of 103 cm~3, coagulation is unimportant within
the sample or during the buoyant rise of the plume. During the
long period the plume spreads and mixes with atmosphere there
is time for coagulation but this is somewhat compensated for
by the low values of (n) brought on by dilution.
Condensation. An indication of the likelihood that condensa-
tion occurs is the degree of saturation for condensable species.
The degree of saturation (S) is defined as
species partial pressure at a temperature Tm _
species saturation partial pressure at a temperature Tm
c - X P
"
Psat
If the value of S is equal to or above unity then condensation
can be expected.
The mole fraction of species j in the diluted sample is
(4)
where r is the dilution ratio and Ms and Ma the sample and air
molecular weight. The saturation partial pressure at a tempera-
ture Tm can be computed from the Clausius-Clapeyron equation,
124
-------
assuming Trouton's rule
h£ = 10.5 R T.
fg b
(5)
in which the heat of, vaporization (hfg) is assumed constant and
estimated at the boiling temperature TT^) at one atmosphere and
assuming that the vapor behaves as an ideal gas and R is the
gas constant. The relationship between the saturation pressure
at Tm and the boiling temperature T^ at one atmosphere can be
expressed as
(1/Psat> =
(6)
If the sample and the dilution air are ideal gases with constant
specific heats and mixing is adiabatic, the temperature of the
diluted sample Tm leaving the DEP can be written as
T + r T,
T = __s d
m (r + 1)
Thus the degree of saturation of species j is
X.
Sj = -^M
XM
(7)
exp
r + 1) 1_"|
s + rTa " TaJ
(8)
The saturation ratio is not easily computed for the plume.
During the plume's buoyant rise there is little mixing but con-
siderable cooling and during the downwind dispersion of the plume
the reverse is true. In addition, there are distinct concentra-
tion gradients in the spreading plume and hence the saturation
ratio will vary with respect to x, y, and z. For these reasons
a direct comparison of the saturation ratio of the SSD sample
and the plume is not possible.
The rate at which condensation increases the diameter of
a single drop is given by Fletcher10 as,
D dD
G =
-f = G [S-l-g +^-] f (Re, Pr')
P Dp
Dp
pv
PL
Dh. 2 p M .
1 + fq KV a f
RT K f
(
(
Re
Re
, Pr
,Pr)
')
-1
(9)
125
-------
The constants a and b relate to the effect of surface tension
and presence of a solute and the function f (Re, Pr) accounts
for convection heat transfer. The numerical value of the func-
tion is close to unity. Each particle in a population of small
particles grows in a similar manner, but as the condensing ma-
terial is removed from the vapor phase the degree of saturation
(S) decreases. It will also change because of cooling and dilu-
tion. To model condensation for a population of particles re-
quires one to know the rate of cooling, and the original particle
size distribution, mass concentration, and saturation ratio.
The computation is complex and was not performed for either the
plume or the SSD sample. The solution is doubly difficult for
the plume since the degree of saturation (S) and temperature
vary with respect to x, y, and z.
While the rates of cooling and dilution in the SSD sample
and plume are not identical, the SSD system more closely approxi-
mates the plume than the EPA Method 5 system in which there is
no dilution and the EPA Method 17 system in which there is neither
cooling nor dilution.
Adsorption. If the diameter of the particle does not change
rapidly, the rate of change of the mass of an adsorbed species
is equal to the rate at which material diffuses from the gas
phase, viz.
dc
s
[c - c 3
dt ~ * il-v v,ej (10)
where cm o and cs are the concentrations of a material per unit
mass and surface area of the particle and where cv refers to
the concentration of the material per unit volume of gas. The
constant K is the mass transfer coefficient. From the theory
of adsorption, equilibrium conditions can be described by the
Freundlich equation, in which cv e is the concentration of the
material in the gas phase at equilibrium.
c = p a c P (11)
v,e p s
The values of a and 6 depend on the molecular species in
question and the units that are chosen. The quantity a is pro-
portional to the temperature while 3 depends on the nature of
the van der Waals forces on the particle's surface.
While the above expression was not solved for the plume
or the SSD sample, it is possible to determine whether adsorp-
tion in the SSD sample is comparable to that which occurs in
126
-------
the plume. Values of 3 are in general integers larger than unity,
Assume for discussion purposes that g=2.0. If the particle does
not initially contain any of the adsorbate, the surface concen-
tration at any time t can be expressed as
c - c (t)
'6 S
c + G(t)
s,e s
where
= 2
T = 2K
(13)
The parameter T is a time constant, such that if the elapsed
time is equal to T the surface concentration will be 47% of its
equilibrium value, or if equal to 2 T the surface concentration
will be equal to 75% of its equilibrium value. The numerical
value of T depends on the mass transfer coefficient K, and the
absorbate mass concentration cv. The value of T is not easy
to compute since it depends on the fluid mechanics of mixing
within the DEP and the plume. If T is considerably smaller than
the residence time of the SSD system, then the surface concentra-
tion will achieve its equilibrium value by the time the sample
leaves the SSD system.
Filter accumulation error — Inherent in any filtration system
is the constant exposure of the collected particles to the gas
stream. Particles collected early in the sampling period will
be exposed to the gas stream longer than those collected at the
end of the sampling period. Errors which may occur will be
called accumulation errors. If the processes of adsorption and
condensation are completed before the sample passes through the
filter, then there will be no accumulation error and the col-
lected particles will represent those in the sample stream.
If however this is not true, and the particles continue to adsorb
and, or condense gas phase species, then the collected particles
will not be representative of those in the sample stream.
An accumulation error may seriously affect studies in which
condensation in the source is being investigated, if the degree
of saturation at the filter is near unity. Under such conditions
condensable vapors in the gas will continually condense on all
particles collected up to that constant. Such an integrative
effect will suggest condensation that graatly exaggerates what
actually occurs in the plume.
CONCLUSIONS
A new source sampling system has been designed that is easier
to use than a conventional EPA Method 5 system. The heart of
the system is a Dilution Ejector Pump (DEP) that uses dry atmo-
spheric air to withdraw a sample from a process gas stream and
simultaneously dilute and cool it with atmospheric air. Within
127
-------
the DEP the sample reaches equilibrium with the dilution air
in a way that replicates the way actual plumes reach equilibrium
with the atmosphere. The particles in the sample are cooled
to approximately ambient temperature and are deposited on a filter
for analysis.
Tests have been conducted in which source samples were simul-
taneously withdrawn by the dilution sampling system and an EPA
Method 5 system; tests were also made with the dilution sampling
system and an in-stack system. Comparison of the particles ob-
tained by the three systems reveals the following.
(1) The dilution sampling system records larger particle
mass concentrations than an EPA Method 5 system.
(2) The size distribution of the particles obtained with
the dilution sampling system is different from that
obtained with an in-stack impactor. Depending on the
combustion process, the sample obtained with the dilu-
tion system indicates the presence of submicron par-
ticles emitted to the atmosphere that are not present
at stack gas temperature.
(3) The presence of various elements on, or in, the par-
ticles is significantly affected by the techniques
with which the sample is obtained.
ACKNOWLEDGEMENT
This research was supported by Grant No. R803560 from the
Environmental Protection Agency administered through the Center
for Air Environment Studies of The Pennsylvania State University
(CAES No. 521-78).
REFERENCES
1. Heinsohn, R.J., J.W. Davis, G.W. Anderson, and E.A. Kopetz,
Jr. The Design and Performance of a Stack Sampling System
with Dilution. Paper No. 76-37.3, Annual Meeting, Air Pol-
lution Control Association, 1976.
2. Heinsohn, R.J., J.G. Wehrman, J.W. Davis, and G.W. Anderson.
A Comparison of the Particulate Matter Obtained Using a
Dilution Sampling System and a Method 5 Sampling System.
Paper No. 77-12.1, Annual Meeting, Air Pollution Control
Association, 1977.
3. Heinsohn, R.J., and J.W. Davis. Design of Stack Sampling
System with Dilution. Center for Air Environment Studies,
Pennsylvania State University, Report No. 494-78, 1978.
128
-------
4. Wehrman, J.G., R.J. Heinsohn, J.W. Davis, and G.W. Anderson.
Instruction Manual for a Stack Sampler with Dilution. Center
for Air Environment Studies, Pennsylvania State University,
Report No. 477-77, 1978.
5. Hedges, K.R., and P.G. Hill. Compressible Flow Ejectors.
Part I-Development of a Finite Difference Flow Model. Paper
No. 74-FE-i, Annual Meeting, American Society of Mechanical
Engineers, 1974.
6. Hedges, K.R., and P.G. Hill. Compressible Flow Ejectors.
Part II-Flow Measurements and Analysis. Paper No. 74-FE-2,
Annual Meeting, American Society of Mechanical Engineers,
1974.
7. Razinsky, E., and J.A. Brighton. Confined Jet Mixing for
Nonseparating Conditions. J. Basic Eng., Trans. ASME 93(4):
333-347, 1971.
8. Hidy, G.M., and S.K. Friedlander. Vapor Condensation in
the Mixing Zone of a Jet. AIChE J. 10(1):115-124, 1964.
9. Slade, D.H., ed. Meteorology and Atomic Energy, 1968.
TID-24190, Environmental Science Services Administration,
Silver Spring, MD, July, 1968. p 192.
10. Fletcher, N.J. The Physics of Rainclouds. Cambridge Uni-
versity Press, New York, 1969. pp 122-136.
129
-------
PAPER 8
AEROSOL CHARACTERIZATION WITH A QUARTZ CRYSTAL
MICROBALANCE CASCADE IMPACTOR
DAVID C. WOODS
NASA LANGLEY RESEARCH CENTER
AND
RAYMOND L. CHUAN
BRUNSWICK CORPORATION
ABSTRACT
A Quartz Crystal Microbalance (QCM) cascade impactor, de-
veloped by Chuan,1 has been used by NASA to obtain in-situ data
on size distribution, elemental composition, and morphology of
aerosol particles. Aerosols in rocket exhaust plumes have been
characterized using the QCM, in which the data are used for de-
veloping dispersion models for predicting plume behavior as a
function of meteorology and for assessing the environmental im-
pact of the effluents. Volcanic effluents have been measured
in the troposphere and characterized with the instrument to sup-
port studies leading to an understanding of the volcano's geology
and its contribution to atmospheric aerosols. NASA has more
recently utilized the QCM to measure both upper tropospheric
and stratospheric aerosols. In this paper a description of the
QCM is presented, including its integration with several aircraft
which have served as platforms. Also a review of the above mea-
surement efforts is given and selected data are presented and
discussed.
INTRODUCTION
An aerosol sensor, designed by Chuan,1 has been used exten-
sively in research at the NASA Langley Research Center for ob-
taining data on a variety of aerosols. The sensor, referred
to as a Quartz Crystal Microbalance (QCM) cascade impactor, mea-
sures aerosol concentration and size distribution (mass concen-
tration as a function of particle diameter) in-situ. It combines
the use of piezoelectric crystal microbalances for sensing mass
in real time with the technique of inertial impaction, with a
130
-------
cascade of impactors for classifying aerosols by aerodynamic
size. Other methods that have been used for measuring size
distributions and concentrations of aerosols in-situ include
collection by impaction with later laboratory analysis such as
weighing, counting, etc., and single particle light scattering.
All methods, while providing valuable information, have uncer-
tainties associated with their measuring techniques. For ex-
ample, the aerosols collected by the impaction methods may under-
go changes and/or evaporation between the time and location of
collection and the time and location of analysis. Thus, this
is not a truly in-situ method. On the other hand, size informa-
tion obtained from light scattering measurements involves un-
certainties in the knowledge of refractive index and particle
geometry. Since the QCM cascade impactor senses mass in real
time, the problems associated with aerosol changes or evaporation
are avoided and since the size is determined by inertial classi-
fication, refractive index is not a factor. It should be ap-
preciated, however, that the aerodynamic size of a particle
depends on its geometry and mass density. In some cases these
factors may be assumed with reasonable accuracy or may be deter-
mined by post-sampling analysis. Thus, the QCM offers a com-
plementary approach to obtaining aerosol data while avoiding
some of the problems encountered with commonly used techniques.
In addition to the size distribution and mass concentration
data, the size separated aerosol collections can be further ana-
lyzed in the laboratory using scanning electron microscopy (SEM)
with energy dispersive x-ray analysis to determine elemental
composition and particle morphology. These data are of parti-
cular interest to researchers involved in the studies of plumes
from power plants, rocket motors, volcanic eruptions, etc. They
are also important to the understanding of the interplay between
the earth's aerosol layer and solar radiation, through scatter-
ing and absorption, which affects visibility and the earth's
radiation budget and thereby may affect global climate.
We have used the QCM as an airborne sensor on several dif-
ferent aircraft to characterize aerosols in rocket exhaust plumes,
in volcanic eruption plumes and in the upper troposphere and
lower stratosphere at several global locations. Size distri-
bution data and data on elemental composition and morphology
were obtained from each set of measurements. We present herein
a brief description of the QCM sensor and its use on various
aircraft. Some typical data are presented.
Instrument Description and Operation
The airborne QCM sensor (shown in the photograph in Fig-
ure 1) consists of ten impaction stages. Starting with stage 1,
the particle diameters for which the impaction efficiency is
50% (50% cut points) are: 25, 12.5, 6.4, 3.2, 1.6, 0.8, 0.4,
131
-------
Figure 7. The 10-stage QCM cascade impactor for airborne measurements.
132
-------
0.2, 0.1, and 0.05 micrometers (urn) for equivalent spherical
particles with a mass density of 2 g/cm3. The impaction surface
in each stage, instead of being an inactive collector such as
a glass plate, a microscope grid, or filter paper, as is used
with conventional impactors, is an active piezoelectric crystal.
The crystal, usually coated with a thin layer of grease to impede
particle bounce, is a part of an oscillating circuit which senses
the mass of the impacted aerosol by the associated change in
frequency. Therefore, by continuously monitoring the oscillator
frequency in each stage, one can determine the aerosol mass col-
lected as a function of particle size.
Because of the very high sensitivity of the piezoelectric
microbalance (of the order of 109 Hz/g) a volume flow rate of
only 240 mi/min is used for sampling at altitudes below 3 km
where the total mass concentration is on the order of 10 to
100 yg/m3. At altitudes above 3 km where the mass concentra-
tion is on the order of 1 yg/m3 the volume flow rate is 2 il/min.
When sampling in ambient air a sampling time of the order of
a few minutes is usually required for a sensible signal while
a sampling time of only several seconds is required when sampling
in highly concentrated plumes such as rocket exhaust plumes or
volcanic plumes.
When the sensor is flown external to the aircraft (e.g.,
under the wing of the Sabreliner and Queenaire, or on the side
of the P-3), it is placed in an aerodynamically-shaped housing
(shown in Figure 1) which protects the sensor and allows air
to flow around the cascade maintaining the sensor at ambient
temperature. Isokinetic sampling at the required flow rate is
obtained by a specially designed inlet probe. It consists of
two diffusers in series, as shown schematically in Figure 2.
The object is to decelerate the air which enters the probe at
the aircraft speed Uo (approximately 40 m/sec to 200 m/sec depend-
ing on aircraft and altitude) to a speed Us which matches the
air speed entering the impactor. Us is equal to V/A, where V
is the volume flow rate into the impactor and A is the area of
its inlet. Since A is not very large (typically about 0.3 cm2),
to decelerate the air from Uo to Us in a single duct would re-
quire that the probe inlet area be too small and it would be
impractical to handle. The solution is to start with a practical
probe inlet area, Ao, of the order of 0.1 cm2 (equivalent diam-
eter of about 0.3 cm); decelerate by a factor of 10 to 20 along
a small divergence angle diffuser (included angle 15°, to assure
no flow separation); skim the core of the decelerated flow with
a sharp-lipped "skimmer" of area A2, which is a small fraction
of the plenum area Ai, and deliver it to the second diffuser.
The excess flow in the plenum of the first diffuser is vented
out to the atmosphere at the truncated base of the diffuser where
the base pressure, p^, is sub-atmospheric.
133
-------
TANDEM DIFFUSER INLET
OJ
STAGE 1
CASCADE HOUSING
Us = CASCADE INLET FLOW SPEED
V = (U0A0)
A2 *S
A1 A4
Figure 2. A schematic of the airborne QCM inlet.
-------
In the second diffuser the velocity U^ is again decelerated
by a factor of about 10, to match the impactor inlet velocity,
Us, which for the case of V = 2000 m£/min and inlet area of 0.3
cm2 is about 100 cm/s (as compared to the airplane speed of 200
m/s, a factor of 200) . The QCM inlet acts as the core "skimmer"
in the plenum of the second diffuser. The excess flow is vented
out of the base of the second diffuser into the housing which
is in turn vented through its base, at base pressure, p^, less
than ambient, into the atmosphere.
The flow relationship through various stations along the
tandem diffuser is shown below the schematic in Figure 2. By
judicious selection of the various areas it will be found that
the impactor inlet can be readily matched to any specified flight
speed.
In a typical sampling flight the instrument is placed in
its standby mode, in which ambient air is taken through the im-
pactor through a total filter while en route to the sampling
site. Just before a plume is entered, the sampling valve is
moved from the standby position to the sampling position and
the air flows directly into the impactor, bypassing the absolute
filter. When the airplane leaves the plume (or when sufficient
signal is obtained in case one is flying along a long plume)
the valve is returned to standby. The frequency change in each
stage and the time in the plume yields the aerosol mass concen-
tration in each stage. These frequencies are recorded by a
printer in the instrument control unit, so that the raw data
can be analyzed after the flight.
Table 1 summarizes the measurement efforts in which the
QCM cascade impactor has been utilized by the Langley Research
Center. The geographical locations, the types of aerosols mea-
sured, and the aircraft used for the measurements are listed
in chronological order. The following will include discussions
on portions of selected experiments from this table. These dis-
cussions are intended to illustrate the kinds of experiments
in which the QCM may provide useful information and the results
that may be expected.
Solid Propellant Rocket Exhaust
Our initial use of the QCM as an airborne sensor was for
sampling and characterizing particulates in the exhaust plumes
from solid propellant rocket motors (Deltas and Titans) at Ken-
nedy Space Center starting in 1974. For the earlier measurements
we used a single stage QCM impactor2 which was a modification
of one previously used as a ground sampling instrument. It was
mounted in the forward baggage compartment of a Cessna 402 with
protruding isokinetic inlet probes supplying the sample to the
sensor. The single stage impactor provided mass concentration
data and no real-time particle size data. It was replaced in
1977 by a 10-stage impactor which was flown in a rocket plume.3
135
-------
Date
TABLE 1. APPLICATIONS OF THE QCM CASCADE IMPACTOR AT LANGLEY RESEARCH CENTER
Approximate
Aircraft Altitude (km)
Location
Type of Measurement
May 77 Fairbanks, Alaska
May 77 KSC
Aug. 77 KSC
Sept. 77 KSC
Feb. 78 Guatemala
Stratospheric Aerosol
Rocket Effluents
Rocket Effluents
Rocket Effluents
Volcanic Effluents
Tropospheric Aerosol
Sept. 78 Laramie, Wyoming
Nov. 78 Sondrestrom, Greenland Stratospheric Aerosol
Dec. 78 KSC Rocket Effluents
Mar. 79 Dallas, Texas Tropospheric Aerosol
Apr. 79 Natal, Brazil Tropospheric Aerosol
Apr. 79 St. Vincent Island Volcanic Effluents (Soufriere)
May 79 St. Vincent Island Volcanic Effluents (Soufriere)
NCAR Sabreliner 10-13
Cessna 402 0.3-2.0
Cessna 402 0.3-2.0
Cessna 402 0.3-2.0
NCAR Queenaire ~5
NASA Wallops P-3 3-7.6
NCAR Sabreliner 10-13
Cessna 402 0.3-2.0
NASA Wallops P-3 3-7.6
NASA Wallops P-3 3-7.6
NASA Wallops P-3 3-7.6
NASA Wallops P-3 3-5
-------
During a typical sampling flight the aircraft made a series
of passes through the center of the visible plume flying in a
figure eight pattern while at the same time changing altitude
between passes. Figure 3 (from data taken in December 1978)
shows a size distribution plot of AC/AlogD as a function of
D, where C is the mass concentration in units of yg/m3, and D
is the particle diameter in micrometers. The bimodal distri-
bution, typical of rocket exhaust plumes, has a sub-micron peak
at O.l-jum diameter and a second peak between 3.2 and 6.4 urn.
70
60
50
40
01
3.
o
I 30
o
20
10 -
J_
0.1
1.0
DIAMETER, j
10
100
Figure 3. Particle size distribution in the exhaust plume from a Titan III rocket
launched at Kennedy Space Center on December 13, 1978.
137
-------
Aluminum oxide (A12O.,) spheres, formed by combustion of the
rocket propellant, have been found to make up the major portion
of the particulates in the large size moded as illustrated in
Figures 4a and 4b. Figure 4a shows an SEM photograph of par-
ticles collected in stage 5 of the cascade impactor corresponding
to 1.6-ym diameter. Figure 4b shows an x-ray energy mapping
for aluminum (Al), where the bright spots indicate the presence
of Al. These spots map into the particles in Figure 4a indi-
cating that they do contain Al. Since they are spherical in
shape and show no other elements in the x-ray energy spectrum
they are presumed to be Al,03 particles. In general, the par-
ticles in the 0.1-ym size mode are more complex, consisting of
single spherical A1203 particles, and agglomerates which contain
sodium, aluminum, sulfur, chlorine, potassium, calcium, iron
and zinc.3 It is suspected that the A1203 particles formed in
the exhaust cloud are formed by two processes thus producing
the bimodal distribution. The spherical particles in the large
size mode are believed to be formed molten alumina and the ones
in the submicron size mode are believed to be formed from gas
phase condensation.
Stratospheric Aerosols
In preparation for a series of intercomparative experiments
with the QCM and several other types of aerosol sensors, the
QCM was recently integrated and test flown on the NASA Ames U-2
aircraft. The U-2 is capable of reaching altitudes of greater
than 20 km. The QCM, its control unit, and printer were mounted
inside an area on top of the fuselage. A stainless steel inlet
probe protruding through a cover plate was used to provide the
sample air. A single three-way control switch was provided in
the cockpit for the pilot to perform all operations.
The following data were taken with the QCM aboard the U-2
during a photographic mission over south central California on
June 11, 1979. The measurements were made during level flight
at an altitude of 19.8 km. Figures 5a and 5b show plots of
AC/AlogD as a function of D, for two consecutive 10-minute in-
tervals. These plots show again a bimodal characteristic in
mass concentration: a weaker one at 0.2-ym diameter and a rela-
tively strong mode for the 1.6-ym diameter particles. In addi-
tion, Figure 5a shows an indication of an increase in AC/AlogD
at 0.6 ym.
The SEM photograph in Figure 6 shows a sample of the par-
ticles collected in stage 8 of the cascade impactor correspond-
ing to the peak in the small size range of the bimodal distri-
bution. The majority of the particles appear to be nearly spheri-
cal in shape and are on the order of about 0.2 ym in diameter,
which is the 50% cut point for stage 8. There are quite a few
irregularly shaped larger particles, on the order of 0.5 to 1.0
micrometers. These particles apparently have mass densities
lower than 2.0 g/cm3, the density for which the 50% collection
138
-------
Figure 4. (a) Scanning electron microscope photograph of particles collected in
stage 5 of the QCM cascade impactor for a Titan rocket exhaust plume.
(b) X-r
-------
2.4
2.0
n
E
01
S1
1.6
1.2
0.8
\
\/
-tit
/
/
0.4
0.0
_L
0.1
1.0
DIAMETER,/zm
10
100
Figure 5a. Size distribution plots of stratospheric aerosol particles measured with
the QCM in the NASA Ames U-2 aircraft at 19.8 km over south
central California.
-------
2.4 -
CO
5
S
Q*
o>
O
2.0
1.6
1.2
0.8
! \ I
0.4
0.0
J_
0.1
1.0
DIAMETER,
10
100
Figure 5b. Size distribution plots of stratospheric aerosol particles measured with
the QCM in the NASA Ames U-2 aircraft at 19.8 km over south
central California.
-------
6. Scanning electron microscope photographs of stratospheric particles
collected in stage 8 of the QCM cascade impactor over south central
California at 19.8 km.
142
-------
efficiencies are determined. The combination of density and
shape causes these particles to behave aerodynamically as if
they are 0.2-micrometer diameter spheres with a mass density
of 2 g/cm3.
Three of the particles in Figure 6 were selected for energy
dispersive x-ray analysis, particles A, B, and C. Particle A
is approximately spherical but is greater than 0.5 ym in diam-
eter. B appears to have more of a cubic shape and is on the
order of about the same size as A. C, on the other hand, is
much smaller than A and B, being on the order of about 0.2 ym
and spherical in shape. The type C particles are much more
abundant than the other types. The x-ray energy spectra for
the three particles are shown in Figure 7 along with a background
spectrum. A and B show the same elements while C shows no x-
ray spectrum at all, either because of its small size or because
there are no elements present that can be detected by this x-
ray method. In general, it was found that the morphology of
the particles as well as the composition differed for the various
impaction stages, indicating that the various types of particles
are size dependent.
(A)
(B)
Figure 7. X-ray energy spectra for particles A, B, and C in Figure 6.
143
-------
Volcanic Plumes
The QCM was aboard the NCAR Queenaire when it made a series
of sampling flights over the active volcanoes, Fuego and Santia-
guito in Guatemala, during February and March 1978. "* During
the sampling period, Santiaguito was producing several large
ash-laden eruption clouds per day, some rising to higher than
5 km above sea level. The aircraft started sampling downwind
of the visible plume and made several passes through the plume
as it approached the volcano. Figure 8 shows two size distribu-
tions obtained from QCM data taken on February 28, 1978, (A)
9 miles downwind and (B) 18 miles downwind. In these plots,
C = 97.7
o
0.1 -
0.1 ~ 1.0 10
PARTICLE DIAMETER,
100
(A) SANTIAGUITO FEBRUARY 28, 1978 - 9 MILES DOWNWIND
C = 425.5
0.1 1.0 10 100
PARTICLE DIAMETER, jum
(B) SANTIAGUITO FEBRUARY 28, 1978 - 18 MILES DOWNWIND
Figure 8. Aerosol size distribution in the active plume of Volcano Santiaguito,
February 28, 1978.
144
-------
the mass concentration C^ in stage i of the cascade impactor
was normalized to the total mass concentration C, where C = £ C^.
The Cjyc vs. D plots may be compared with the AC/AlogD vs. D
plots by multiplying C^/C by C/AlogD where C^ is the same as
AC. These particular plots in Figure 8 show three modes in the
size distribution. Many of the distributions, however, showed
only two modes. The plumes were not very uniform and the size
distributions tended to depend on the spatial position through
which the aircraft made its pass. The particles collected from
the plume were observed with an SEM. They exhibit the character
of a crystal-rich volcanic magma consisting of crystal fragments
of plagioclase and magnetite. It is believed that the sub-micron
mode in the size distribution is caused by fragmentation caused
by collisions among the particles. Since the large particles
are crystalline, they can easily be broken into sub-micron size
fragments by collision.
CONCLUSIONS
The basic operating principles of a ten-stage quartz crys-
tal microbalance cascade impactor have been described. It is
designed to obtain real time in-situ data on aerosol size dis-
tribution and to collect aerosols for post-sampling analysis
of their elemental composition and morphology as a function of
size. Examples of data have been presented which demonstrate
the satisfactory performance of the instrument. These data were
not obtainable in the past with previously existing sensors.
There are many potential applications of this instrument for
both tropospheric and stratospheric studies and. it is anticipated
that additional atmospheric measurements will continue to be
made using this sensor.
REFERENCES
1. Chuan, R.L. Rapid Measurement of Particulate Size Distri-
bution in the Atmosphere. In: Fine Particles, Aerosol Gene-
ration, Measurement, Sampling, and Analysis. Benjamin Y.H.
Liu, ed. Academic Press, New York, 1976.
2. Woods, D.C. Measurement of Particulate Aerosol Mass Concen-
tration Using a Piezoelectric Crystal Microbalance. In:
Aerosol Measurements, D.A. Lundgren, et al., eds. University
Presses of Florida, 1979.
3. Woods, D.C. Rocket Effluent Size Distributions Made With
a Cascade Quartz Crystal Microbalance. In: Proceedings,
4th Joint Conference on Sensing Environmental Pollutants,
1977.
4. Rose, W.I., Jr., R.L. Chuan, R.D. Cadle, and D.C. Woods.
Small Particles in Volcanic Eruption Clouds. Am. J. Sci.,
in press.
145
-------
PAPER 9
EXTENDING PRECISION IN A COMPUTER-BASED CASCADE
IMPACTOR DATA REDUCTION SYSTEM
JEAN W. JOHNSON
B.E. PYLE
WALLACE B. SMITH
SOUTHERN RESEARCH INSTITUTE
ABSTRACT
Due to the increased interest in the health effects of
particulates up to 15.0 ym diameter, it has become necessary
to accurately extrapolate the cumulative mass concentration be-
yond the effective limit of the first stage D50 (approximately
10.0 ym) of a cascade impactor. A first order osculating poly-
nomial is proposed in conjunction with the Computer Impactor
Data Reduction System (CIDRS, EPA 600/7-78-042) for fitting the
cumulative mass curve between the first stage Dso and the maximum
particle size. The function is a third degree polynomial which
uses the known characteristics of the cumulative mass curve for
its solution over this range of particle sizes. Testing this
technique on a number of theoretical unimodal and bimodal size
distributions demonstrates a high degree of accuracy in recover-
ing the true, cumulative, particle concentration <15.0 ym. The
technique described is proposed for recovering inhalable par-
ticulate from existing data and until such time that a cascade
impactor is designed with a stage cut of 15.0 ym or greater.
BACKGROUND: THE COMPUTER-BASED IMPACTOR DATA REDUCTION SYSTEM
In March of 1978, the EPA report entitled "A Computer-Based
Impactor Data Reduction System" was released,1 The report de-
scribes a series of five computer programs, known by the acronym
CIDRS, which are designed to reduce the field data taken by com-
mercially available round jet cascade impactors. An outline
of the CIDRS programs is illustrated in Table 1 with a listing
of the program names and their functions.
146
-------
TABLE 1. CIDRS PROGRAMS AND OUTPUT
MPROG
Cumulative mass concentration
-------
particle sizes. Both tabular and graphical outputs are produced
at each of the data reduction levels described.
The curve fit to the cumulative mass distribution is a criti-
cal point in this data reduction system. After the fit is made
in program SPLINl, the stage by stage values of cumulative mass,
the D50's, the stage by stage values of AM/AlogD, and the geo-
metric mean diameters are ignored. Averaging and penetration-
efficiency calculations are based strictly on the fitting coef-
ficients as stored by SPLINl execution.
The method of fitting used in SPLINl is a graphical tech-
nique, or a "programmed French curve". The Iog10 (cumulative
mass concentration) vs. Iog10 (D50) points are input to SPLINl
as illustrated in Figure la. Then a parabola is fitted to each
set of three consecutive cumulative mass points. As shown in
Figures Ib through Id, interpolated points are defined at equal
log diameter intervals between the first two of each set of
three D5p's. (A parabola having a negative first derivative
in this interpolation area is overridden by a straight line
between the two D50's to ensure a continuously increasing cumu-
lative mass curve.) The final interpolation parabola (Figure
le) is fitted through the second D50, first D50l and maximum
particle size DMAX.
In the original version of SPLINl, a hyperbolic function
of the form
M = ai +
is used to define the interpolation points between the first
Dso and DMAX, where M is the mass concentration (mg/acm) less
than particle diameter D(ym) and aj and a2 are constants. Seven
equally log-spaced interpolated points are defined over this
broad particle range. The hyperbolic function as used for point
interpolation is illustrated in Figure If.
The function which is proposed to replace the hyperbola,
a third-degree polynomial of first order osculation, is the sub-
ject of the remainder of this paper. Such a function, used to
interpolate points between the first stage D50 and the maximum
particle diameter, is illustrated in Figure 2. Note that as
the osculating polynomial approaches the total mass loading,
its first derivative goes to zero. For log-diameters greater
than this zero slope point (ZSPT) the cumulative mass loading
is at its total value. The functional form of the osculating
polynomial is discussed below. Its importance lies in the fact
that it is a better suited function than the hyperbola for re-
covering the inhalable particulate concentration, . i.e. , the
cumulative mass concentration < 15.0 urn, which lies in this
fitting region beyond the first stage D50 K10.0 jam) .
148
-------
5
CO
CO
<
2
Ui
D
O
D50(6) D50i5) D50{4) D50(3) D50(2) D50<1)
PARTICLE DIAMETER
Figure la. Cumulative size distribution from raw impactor data.
Z
Q
CO
1
UJ
5
o
DMAX
FIRST INTERPOLATION
PARABOLA
INTERPOLATED POINTS
i i 11
I'M ""
D50(6) D50(5) D50(4) D50(3) D5Q(2) D50(1) DMAX
PARTICLE DIAMETER S4181-7
Figure 1b. Start of development of interpolated points between first and last
149
-------
o
5
<
o
LU
>
D
O
SECOND
INTERPOLATION
PARABOLA
INTERPOLATED POINTS
4-
I i i i i i
I'M
D50(6) D50(5) D50(4) D5Q(3) D50(2) D50(1) DM AX
PARTICLE DIAMETER
Figure 1c. Continued generation of interpolated points
z
Q
LU
D
O
INTERPOLATED POINTS
O
THIRD INTERPOLATION PARABOLA
II I I I I I I I I I I I I I I 11
I I I I I I
i i i i i i
D50<6) D50(5) D50(4) D50<3) D50<2> D50<1)
PARTICLE DIAMETER
Figure Id. Continued generation of interpolated points
DMAX
S4181-8
150
-------
5
CO
co
LU
O
INTERPOLATED POINTS ON
FINAL PARABOLA
FINAL INTERPOLATION
PARABOLA
00'
' , ' I ' ".'
1 I I I I 1
i i I J
o
z
a
<
o
CO
co
O
D50(6) D50(5) D50(4) D50(3) D50(2) D50(1) DMAX
PARTICLE DIAMETER
Figure 1e. Generation of interpolated points on parabola
which includes DMAX.
SLOPE = O
HYPERBOLA AND
HYPERBOLIC
INTERPOLATION POINTS
BETWEEN
D50 (1) and DMAX
j i
•H
D50(5) D50(4) D50(3) D50(2) D50(1)
PARTICLE DIAMETER
i i
DMAX
S4181-9
Figure If. Generation of interpolated points on hyperbola through
D50(1) and DMAX
151
-------
The set of interpolated and original D50 points is then
used in fitting a series of continuous, second-degree polynomi-
als. The fitting coefficients along with their boundary points
(the set of cumulative mass concentration vs. particle size used
in making the fit) are stored in files. Using these coeffici-
ents, the cumulative mass curve and the mass and number size
distribution curves may be recovered for use in any subsequent
program.
EXTENDING PRECISION IN THE MASS SIZE DISTRIBUTION BEYOND THE
FIRST STAGE D50
There is an increasing interest in the health effects of
particulate matter consisting of particles up to 15.0 urn in diam-
eter. The EPA has begun a large program to measure the emissions
of this inhalable particulate from stationary sources. Also,
efforts will be made to recover the inhalable particulate (IP)
concentration from existing data by extrapolating the cumulative
mass concentration curve beyond the first stage Dso limit to
15.0 ym.
2
5
1
5
u
INTERPOLATION
POINTS ON AN
OSCULATING
POLYNOMIAL
INTERPOLATION
POINTS ON LINE
(SLOPE=0) AT
TOTAL MASS
LOADING
il
t>50<6> D50(5) D50<4) D50(3) D50I2) D50(1)
PARTICLE DIAMETER
DMAX
S4181-10
Figure 2. Generation of interpolated points on osculating polynomial and
zero slope line at total mass concentration.
152
-------
CIDRS has proven to be an accurate and time saving tool
for impactor data reduction for the size region of approximately
0.25 pm up to 10.0 ym or the first stage D50.2 Because of the
present construction of cascade impactors, there is a large span
of particle sizes from the first stage cut point up to the largest
particle size in the distribution. Due to the lack of informa-
tion, it is difficult to describe the size distribution in this
range of particle sizes. Since an extrapolation to 15.0 ym ex-
tends the data only slightly beyond the first stage D5„, extrac-
tion of IP mass concentration appears feasible if proper extra-
polation techniques are used.
In view of this an extrapolation method that uses only the
known parameters would seem ideal. The known or obtainable
properties of the distribution are:
1. The first (largest) D5o is known.
2. The cumulative mass at the first D5o is known.
3. The slope of the cumulative mass curve at this Dso can
be calculated.
4. The largest particle diameter is known or can be esti-
mated .
5. The total cumulative mass at the largest particle diam-
eter is known.
6. The slope of the cumulative mass curve at the largest
particle diameter is known (=0.0).
7. The slope of the cumulative mass distribution is non-
negative .
Until now the function used in CIDRS for fitting data beyond
the first stage D50 has been a hyperbola of the form given in
Equation 1.
Properties 1, 2, 4, and 5 are automatically satisfied by
this two point fit (Equation 1). In addition, the form of the
hyperbolic function generates a close approximation to properties
6 and 7 in that the slope of this function for real data is never
negative and approaches zero at large D.
However, the hyperbolic function has drawbacks which limit
its usefulness for extrapolation to inhalable particle sizes
above the first stage Dso. First, property 6 is never rigorously
satisfied. While the slope of the hyperbolic function approaches
zero for large values of the maximum particle size, it will not
be zero for a finite maximum particle size as is required. More
seriously, there is no correlation between the slope of the hyper-
bola at the first stage D50 and the slope determined by the spline
153
-------
fitting routine. This discontinuity in slope violates property
3 of the true cumulative mass distribution and thus the hyper-
bolic function will not define proper mass concentrations for
diameters slightly above the first stage Dso.
THE OSCULATING POLYNOMIAL
The new technique employs a polynomial of first order oscula-
tion to fit cumulative mass concentration from the first stage
D50 to the maximum particle size. It not only passes through
the two end points, but also may be constructed to have zero
slope at the maximum particle size and have a slope at the first
stage D50 equal to that of the spline fit through the first stage
D50. The osculating polynomial may also be constrained to have
a non-negative slope over the specified range of particle sizes.
In other words, all seven of the known or obtainable properties
of the mass distribution in this particle range (as listed in
the preceding section) can be satisfied by the proposed technique,
A polynomial of first order osculation is used as an approxi-
mation to a given function. By definition it matches both the
function and its first derivatives at a finite number of points.3
In general, there exists a unique (2n-l) degree polynomial fit
to a set of n points. For the case of impactor data, let par-
ticle diameter be represented by the variable D, the cumulative
mass loading by M(D) , and the osculating polynomial approximation
of this function by P(D); then properties 1-6 require
P(Di) = M(Di), and P'(Di) = M'(Di) (2)
for i = 0,1 where the first point corresponds to the first stage
D50 and the second to the maximum particle diameter. For the
approximating equation to be physically realistic it is also
necessary that property 7 be satisfied; therefore
P'(D) _> 0 for D0 <_ D _< D:. (3)
To first order osculation, the polynomial P(D) can be de-
scribed by Hermite's formula,
n n
P(D) = £ U. (D)M(D.) + £ V. (D)M'(D.) (4)
i=0 L x i=0 1 1
where M(Di) and M1 (Di) are the known values of mass concentra-
tion and its derivatives at D0 and D^- thus n has values of 0
and 1 only. The functions Ui(D) and Vj.(D) can be expressed in
terms of the Lagrange multipliers Li(D) and their derivatives
Li1 (D) such that
" 2 (5)
(6)
U^D) =j"l-2Li'(Di) (D-D^lfL
154
-------
where
n (D-D.)
(7)
i = 0,1,2 , ---- ,n.
For the two points i = 0, i = 1 we have from Equation 7
_
Ll(D) = L- ,9)
DI - DO
and their derivatives are
= U0>
For the points i = 0, i = 1 Equation 4 becomes
P(D) =U0(D)M(D0) +U1(D)M(D1) +V0(D)M'(D0) + Vj (D)M' (D! ) . (12)
Combining Equations 5 and 6 with 8, 9, 10, and 11, the osculat-
ing polynomial then becomes
+ I"(D - Dp) (D - DI) 2 M' (D0)1
L (D0 - Dx)2 J
' (Di)1
J'
- D,)
- D0)
(13)
Equation 13 can be put into a simpler form by defining the follow-
ing constants:
155
-------
M(D0)
a2 =
a3 =
a,. =
(D0 - D,)
2M(D1)
(D, - D0)
3'
2M(D0)
~~"
(D0 - D,)3
M'(D0)
(D0 - D,)
2'
(D, - D0)
(D, - D0)
= k
[kx + k3 - (2 Dj + D0) (k2 + ks) - (2D0 + D^
[(k2 + ks) (D? + 2 DoDj) + (k,, + k6) (D2 + 2D0
- 2kiD! - 2k3D0]
[k^2 + k3D2 - D0D2(k2 + ks) - D2Dt (k, + k6)]
k6)]
In terms of these constants, the first order osculating poly-
nomial can be written as a cubic equation in particle diameter
D such that
P(D) =
and its first derivative is
+ a2D + a3D +
P1 (D) = 3ajD2 + 2a2D + a
3.
(15)
(16)
Property 7 above for impactor data limits the acceptable
solutions 15 and 16 to those for which
P1 (D) >_ 0, D0 £ D 1 Dx. (17)
For some combinations of impactor data, particularly those
for which the slope M' (D0) of the cumulative mass distribution
at the first D50 is large, property 7 places an upper limit on
the range (D0 to Dx) of permissible particle diameters. For
those cases, physically acceptable solutions can be found for
a reduced range of particle diameters wherein the maximum size
is somewhat smaller than the original Dj or assumed maximum par-
ticle size. Figure 3 illustrates this. The first fitting oscu-
lating polynomial has the proper cumulative mass loading value
at the first D50 and at the maximum particle diameter. The first
derivative of the function at these two particle sizes is also
correct, i.e., the first derivative of the osculating polynomial
with respect to log diameter is the same as that for the spline
fit at the first D50 and is zero at the maximum particle size.
However, since the osculating polynomial does have negative first
derivative values for D0
-------
tested for negative first derivative values for Do£D£Di. The
particle size DI continues to be redefined at smaller values,
and the osculating polynomial is fit between Do and DI until
DI is defined at a critical value such that there are no negative
derivatives over the tested particle size range for the oscu-
lating polynomial. DI, renamed "zero slope point", ZSPT, marks
<
O
CO
CO
D
O
1ST FIT (THROUGH
ORIGINAL DMAX)
FINAL (CRITICAL) FIT
DO
(D50(D)
(ZSPT)
(DMAX)
PARTICLE DIAMETER
S4181-11
Figure 3. Fitting the cumulative mass distribution for DQQ(1)<^D0 for DQ
-------
the maximum of the osculating polynomial at the total cumulative
mass loading. As illustrated in Figure 2, interpolation points
for the cumulative mass distribution are then defined along the
osculating polynomial for D50(1)£D£ZSPT and are defined as the
total cumulative mass loading for" 'ZSPT
-------
error in IP recovery may be caused by assuming a largest particle
diameter (100.0 jim here) which is too small with respect to the
20.0 vim MMD.
Variation of the Geometric Standard Deviation of the Particulate
Mass Distribution
To test the sensitivity of this fitting technique to dif-
ferent values of the aerosol geometric standard deviation, og,
unimodal particle-size distributions having Og values of 1.5
to 3.5 were tested. Figure 5 shows the ratio of recovered to
true inhalable particulate IPR/IPT vs. these values of o~ for
stages having collection curves with geometric standard devia-
tions of 1.3 and 1.06. The technique shows good accuracy in
recovering the true IP concentration. The highest ratio value
here is 1.04 for a relatively sharp mass distribution with stand-
ard deviation of 1.5. The accuracy increases as the particle-
size distribution broadens.
Q.
HI
D
OC
t—
i
VERED
O
LU
OC
I.UO
1.05
1.04
1.03
1.02
1.01
0.99
I I I ? " ~ ' "
ASSUMED DMAX = 100.0 Aim
_ og = 2.5
LEGEND
0 crgs = 1-3
• ags = 1.06
— —
_ •
IB
-OB
[H
i . • i
1.0 2.0 5.0 10.0 20.0 100.0
MASS MEDIAN DIAMETER (MMD), Aim S4181 12
Figure 4. Ratio of recovered to true inhalable particle concentration versus
mass median diameter of a unimodal log-normal particle size
distribution
159
-------
Variation of the Geometric Standard Deviation (Slope) of the
Collection Efficiency Curve
The geometric standard deviation of the collection effi-
ciency curve of the sampler, Ogs' ranges from a near perfect
cut value of 1.01 to 1.7 in this test. As shown in Figure 6,
the variation of OgS has little effect on an accurate recovery
of the inhalable particle concentration. The maximum error for
IP recovery occurs for a sampler efficiency ags of 1.01. Here
the recovered over true inhalable particulate ratio, IPR/IPT,
is only 1.02.
Variation of the Assumed Largest Particle Diameter in the Distri-
bution
In this test the assumed largest particle diameter, DMAX,
is varied from 20.0 ym up to 999.0 ym for the same unimodal log-
normal size distribution. The results, as seen in Figure 7,
show that accurate IP recovery is relatively insensitive to as-
sumed DMAX unless it is extremely undervalued. In this distri-
bution the ratio of recovered to true inhaiable particulate
IPR/IPT has its highest values of 1.07 (for ags = 1.06) and 1.06
(for aqs = 1.3) for an assumed DMAX of 20.0 ym. However, for
this size distribution having a mass median diameter of 5.0 ym,
2= 1.04
in
cc
i-
&
o
LU
CC
LU
O
O
111
CC
1.03
1.02
1.01
1.00
0.99
I I I
ASSUMED DMAX = 100.0
MMD = 5.0 urn
LEGEND
O ags
1.3
1-06
I
I
I
J_
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
GEOMETRIC STANDARD DEVIATION, ag S4181 13
Figure 5. Ratio of recovered to true inhalable particle concentration versus aerosol
geometric standard deviation of a unimodal log-normal particle size distribution.
160
-------
UJ
0
HI
DC
UJ
o
o
UJ
CC
1.02
1.01
1.00
1 1
A
_ A
1 I
1 1 1
ASSUMED DMAX = 100.0 jum
MMD = 5.0 urn
ag=2.5
A A A
1 1 1
1.01 1.06 1.3 1.5 1.7
STAGE EFFICIENCY STANDARD DEVIATION, ags
S4181-14
Figure 6. Ratio of recovered to true inhalable particle concentration versus stage
efficiency standard deviation using a unimodal log-normal particle-size
distribution.
I.U/
1.06
1.05
o.
"J 1.04
cc
oT
RECOVERED I
b b
rs) w
1.01
1.00
0.99
1C
M III I
MMD = 5.0 nm
ag=2.5
LEGEND ~
0 ags = 1.3
• ags = 1-06 -
: . . ;
a a o
i iii i
,
1.0 20.0 50.0 100.0 200.0 500 1000.0
ASSUMED LARGEST PARTICLE DIAMETER, DMAX, /im S4181-15
Figure 7. Ratio of recovered to true inhalable particle concentration versus assumed
largest particle diameter using a unimodal log-normal particle-size
distribution.
161
-------
an assumed DMAX of 20.0 ym is unreasonably low. The other I
values are on the order of 1.01 to 1.02. It is concluded then
that only an extremely low guess of DMAX might cause large errors
in IP recovery. This data reduction method still produces good
IP recovery if the assumed DMAX is overapproximated.
TESTING WITH BIMODAL PARTICLE-SIZE DISTRIBUTIONS
For bimodal testing a log-normal mass distribution is used
having mass median diameters of 2.0 ym (MMDX) and 15.0 ym (MMD2)
and geometric standard deviations, ag and og , of 2.0 for each
mode unless otherwise specified in the text. 2As in unimodal
testing the masses are collected using stage efficiency standard
deviations, ags, of both 1.06 and 1.3 to simulate greased sub-
strates and glass fiber substrates, respectively. Also, unless
otherwise specified, the assumed largest particle diameter, DMAX,
is input as 100.0 ym, and there is an equal contribution of mass
from each of the two modes.
Variation of Geometric Standard Deviation of the Particle-Size
Distribution
In this test, the geometric standard deviations of each
mode of the particle-size distribution, agi and ag , were varied
from 1.5 to 3.0. As seen in Figure 8, the recovery of inhalable
m
- 1.00
DC
UJ
8 0.95
UJ
oc
0.90
0.85
T
ASSUMED DMAX = 100.0 urn
MMD-! = 2.0 Mm MMD2 = 15.0 ^m
FRACTIONAL CONTRIBUTION OF MODES: 0.50/0.50
LEGEND
O ags = 1.3
• ags = 1.06
I
I
1.0
1-5 2.0 2.5 3.0
GEOMETRIC STANDARD DEVIATION (ag1 = ag2)
S4181-16
Figure 8. Ratio of recovered to true inhalable particle concentration versus aerosol
geometric standard deviation of a bimodal log-normal particle-size
distribution.
162
-------
particulate appears more sensitive to this change of parameter
than any other test. For a stage efficiency ags of 1.06, the
recovered IP is as much as 14% low for a distinctly bimodal dis-
tribution where geometric standard deviations ag1 = ag2 = 1.5.
As the modes broaden, the recovered to true inhalable particulate
ratio IPR/!PT approaches 1.0. This should be expected since
the degree of curvature decreases in the cumulative mass distri-
bution as these modes broaden; i.e., the distribution becomes
more unimodal and the spline fitting technique must make less
drastic attempts to adjust to curvature.
Variation of the Mass Median Diameter of the Second Mode
As seen in Figure 9, variation of the mass median diameter
of the second mode, MMD2, has little effect on the accuracy of
IP recovery. The recovered to true inhalable particulate ratio,
IPR/IP-T, remains at or less than 1.05 as the MMD2 ranges from
5.0 ym to 30.0 ym. The highest value of IPR/IP? of 1.05 occurs
for an MMD2 of 30.0 ym. This may be attributed to the large
amount of mass in the second mode which falls beyond the first
1.05
LLJ
D
DC
CC
LLI
o
o
LLJ
CC
1.00
0.95
I I I 1^
ASSUMED DMAX = 100.0 (tm
MMD-] = 2.0 jum
ag1 = ag2 = 2-°
FRACTIONAL CONTRIBUTION OF MODES: .50/.50
LEGEND
O ags = 1.3
B ogs=1.06
I
I
I I
I
1.0 5.0 10.0 15.0 20.0 30.0
SECOND MODE MASS MEDIAN DIAMETER, MMD2, Aim
100.0
S4181-17
Figure 9. Ratio of recovered to true inhalable particle concentration versus second
mode mass median diameter of a bimodal log-normal particle-size
distribution.
163
-------
Variation in Fraction of the Total Mass Contained in Each of
the Modes
In this test the ratio of mass contributed by each of the
two modes in the bimodal distribution was varied. Beginning
with a 25% contribution of mass in the first mode and a 75% con-
tribution of mass in the second mode, the ratio was changed on
each trial with a greater fraction of mass being contributed
by the first mode in each successive trial. Figure 10 shows
that the recovery of inhalable particulate less than 15.0 ym
of this bimodal distribution is acceptable regardless of the
ratio of modes. The ratio of recovered to true inhalable parti-
culate IPR/IPT is at its lowest value of 0.931 for glass fiber
substrates, where the ratio of first mode mass to second mode
mass is 0.25/0.75. As mass becomes less concentrated at sizes
greater than the first stage D5o, i.e., as the mass mode ratio
approaches 0.75/0.25, IPR/IPT approaches a perfect 1.00.
1.00
a.
UJ
cc
a
HI
cc
£ 0.95
O
0
111
cc
0.90
II I I
O
0 •
g ASSUMED DMAX
O MMDT = 2.0 jum
0 ogl = 2.0
I
•
= 100.0 Mm
MMD2= 15.0 Mm
ag2 = 2.0
•
LEGEND
0 ogs = 1.3
• a
• UgS
II I I
0.25 0.35 0.50 0.65
0.75 0.65 0.50 0.35
= 1.06
I
0.75
0.25
FRACTIONAL CONTRIBUTION OF FIRST MODE,
FRACTIONAL CONTRIBUTION OF SECOND MODE, F2
S4181-18
Figure 10. Ratio of recovered to true inhalable particle concentration versus fractional
contribution of mass from each mode of a bimodal log-normal particle-
size distribution.
164
-------
CONCLUSIONS
It is concluded from the test results discussed in this
paper that, for data taken by a round jet cascade impactor, a
polynomial of first order osculation may be used for fitting
the cumulative mass concentration curve between the first stage
D50 and some maximum sampled particle size. This function used
in conjunction with the "Computer-Based Cascade Impactor Data
Reduction System"1 provides a method for accurately determining
the inhalable particulate (IP) concentration, i.e., the cumu-
lative mass concentration of particles _<15.0 urn in diameter.
The method does not require the use of a 15.0 pm cut diameter
sampling device. It therefore may be used for determining IP
concentration in gas effluents sampled to date by conventional
impactors.
To test the accuracy of the osculating polynomial for de-
termining IP concentration, a number of unimodal and bimodal
log-normal size distributions were theoretically sampled accord-
ing to the impactor's stage efficiency curves and the resulting
IP concentration was compared to the original values. Varia-
tions of the mass median diameter and geometric standard devia-
tion of the size distributions, the assumed largest particle
diameter, stage efficiency standard deviation, and fractional
contribution of mass per mode (in the case of bimodal distri-
butions) were studied as to their effect on recoverable IP con-
centration parameters.
Good recovery of the IP concentration was obtained for all
"sampling runs", within approximately ± 7%, except in the case
of sampling a bimodal distribution where the modes are very
sharply defined and with well separated mass median diameters.
Specifically, in sampling a bimodal distribution with mass median
diameters of 2.0 pm and 15.0 urn with aerosol geometric standard
deviations of 1.5 for each mode, the recovered IP concentration
was 14% below the true IP concentration. The greater degree
of error here is attributed to the sharp changes in curvature
of the cumulative mass concentration curve to which the modified
spline fitting procedure must adjust. It can reasonably be
assumed that actual field data would fall into size distribution
modes such that the dispersion of each mode could be described
by a geometric standard deviation value of 1.5 pm or larger.
Therefore, the degree of error in recovering IP concentration
using this curve fitting procedure may be expected to be within
the limits of ± 14%.
ACKNOWLEDGEMENTS
This research was supported by the U.S. Environmental Pro-
tection Agency under Contract No. 68-02-3118, D.Bruce Harris,
Project Officer. Joseph D. McCain and Dr. Ashley D. Williamson,
at Southern Research Institute, provided information and helpful
criticism of the paper.
165
-------
REFERENCES
1. Johnson, J.W., G.I. Clinard, L.G. Felix, and J.D. McCain.
A Computer-Based Cascade Impactor Data Reduction System.
EPA-600/7-78-042, U.S. Environmental Protection Agency,
Research Triangle Park, NC, 1978. 592 pp.
2. McCain, J.D. A Data Reduction System for Cascade Impactors,
EPA-600/7-78-132a, U.S. Environmental Protection Agency,
Research Triangle Park, NC, July 1978.
3. Scheid, Francis. Numerical Analysis. Schaum's Outline
Series, McGraw Hill, New York, 1968. p. 65.
166
-------
PAPER 10
IMPLEMENTING CIDRS - A PROGRAMMER'S PERSPECTIVE
CLINTON E. TATSCH
RESEARCH TRIANGLE INSTITUTE
QUALITY ASSURANCE: THE TOTAL PERSPECTIVE
As we work with research quality assurance, it is clear
that investigators in various disciplines are becoming aware
that an appropriate quality assurance program is not an add-on
or a separate function that can be neatly applied to existing
programs. It is increasingly clear that quality assurance can
only be planned and performed in any adequate sense by indivi-
duals who are personally committed to producing high-quality
data. This commitment is demonstrated by a concern, in the early
planning stages of a project, that estimates of precision and
accuracy be obtained for each measurement or observation. By
characterizing data quality as the data are collected, it becomes
possible to estimate the quality of the final results. The most
simple to estimate are relatively straightforward measurements
such as length or weight. These measurements can be made re-
peatedly and carefully on everyday objects using physically
"comfortable" procedures. At the other end of the scale, the
assessment of computer software quality, by its very nature,
is difficult to thoroughly check. The purpose of the computerized
cascade impactor data reduction system (CIDRS) is to perform
a series of lengthy calculations that cannot be done manually
due to time constraints. Additionally, as a result of having
this capability of performing more calculations more quickly,
computer programs are becoming much more intricate in their
decision-making (i.e., branching) capabilities; as this occurs
it becomes more difficult to thoroughly check the validity of
all branching options. However, work is in progress along these
lines. We believe our work with the CIDRS system in the past
few months has been useful in uncovering some of the pitfalls
in the area of software quality assurance.
167
-------
Our objectives in implementing CIDRS are:
Modify and implement CIDRS coding for the computer in-
stallation at Research Triangle Institute
Execute documentation examples
Begin sensitivity analysis
• Make changes in such a way as to be most useful to other
users
The first three were in the task directive from EPA: we com-
pleted the first two, and began work on the third objective.
The fourth objective, that of making changes in the fashion that
we did, developed shortly after we began work, and eventually
assumed a place of major importance.
IMPLEMENTATION: AN OVERVIEW
In order to make any evaluation of CIDRS, it had to be exe-
cutable on our computer. CIDRS was developed on a PDF 15/76
over a period of several years; hence the goals and intent of
the system have changed significantly since the early code was
written. In order to execute CIDRS programs, the source code
must be readable and compatible with the resident FORTRAN com-
piler. The control language used must also function on the local
computer system.
With this in mind, we obtained a tape containing CIDRS from
Southern Research Institute and began its implementation on the
IBM 370/168 at the Triangle Universities Computing Center (TUCC).
CIDRS is made up of six mainline programs as shown in Figure 1.
These programs run independently of each other with the excep-
tion of the generated data files, which are used by various pro-
grams of the system.
As indicated in Figure 1, MPPROG takes the raw impactor
run data (flow rate, temperature, gas composition, stage load-
ings, etc.) and generates data file KMC001, which contains cumu-
lative mass distribution data. For each impactor run, these
data are used in the curve-fitting program SPLIN1 to generate
FILSPL, which contains the fitting coefficients discussed earlier
These two physical files may contain data for up to 50 impactor
runs with the only stipulation being that the same type of im-
pactor must have been used for each run on the file. These two
files, KMC001 and FILSPL, then serve input to GRAPH, which pro-
duces plotter output for visual checks on the individual or
combined runs, or as input to STATIS, which then averages the
series as into one set of interpolated points with confidence
limits corresponding to all the runs on the file. The stipula-
168
-------
9«vlc« OUTLET
Raw Data
\
/
Printer
Output
Fi«n\J g^j I/
Figure 1. CIDRS program relationships.
169
-------
tion embedded in the logic of the system is that all the impactor
runs included in the KMC001 and FILSPL data files should be
logically related to process conditions. This may conclude the
calculations. However, if a pollution control device is being
evaluated, the data for the associated inlet and outlet measure-
ment series are processed independently, starting at MPPROG.
One series of data files then relates to data collected at the
inlet of the control device; the other series relates to data
collected at the outlet of the control device. After the data
are separately averaged in STATIS, the JWJ001 and JWJ002 files
are used by PENTRA and PENLOG to calculate collection efficiency
(or penetration) for the control device involved.
As indicated in Figure 1, the first two programs in the
series, MPPROG and SPLINl, use direct access files and the system
card reader and printer, whereas the last four, GRAPH, STATIS,
PENTRA, and PENLOG, use the system plotter as well as direct
access files and the system printer. In the case of CIDRS,
plotter usage is proving to be the major source of trouble in
implementing the system. The protocols for addressing peripheral
equipment from within an executing program vary not only from
manufacturer to manufacturer, but from installation to installa-
tion, due to the varying local perspectives.
The basic difficulties one might expect to encounter in
implementing CIDRS would be:
1. Differences in the specific version of FORTRAN used
from one machine to another, and
2. Local idiosyncrasies.
CIDRS: OUR EXPERIENCE
The basic approach we took to implementing the CIDRS system
was to modify the original source code as little as possible,
using the IBM Utility IEBUPDTE to insert only the necessary
changes for each run. In this way, we developed a set of six
modification packets, as indicated in Figure 2, which are applied
in a two-step fashion, thus leaving the original source code
intact and consistent with the documentation. We took this
approach rather than physically modifying the original source
code because in most cases the same types of changes will need
to be made at other installations using, for example, a Univac
or CDC machine, and leaving the modifications intact will high-
light necessary changes for subsequent applications programmers.
We also used the WATFIV interpreter for most of the initial work,
which meant that we executed from source code every time rather
than going from object modules, thus getting more thorough diag-
nostics .
170
-------
Editing
File
"Original"
SoRI Source
File
MASTER
Edited,
Compilation-ready
File
compile, etc,
Figure 2. Structure and use of modifications to original CIDRS logic.
The difficulties we encountered in implementing CIDRS on
our computer system were all installation-specific problems and
for the most part could be solved in a very straightforward but
laborious way. One problem we encountered was the difference
between IBM and Digital Equipment protocols. A second, more
frequent problem involved the I/O handlers.
The hardware-related problem that we identified was that
the PDF 15/76 apparently clears core after each job, whereas
the IBM 370/168 does not. This showed up in various places as
uninitialized variables. Since the IBM FORTRAN-G compiler does
not easily check for uninitialized variables, out-of-bounds array
addresses, etc., the WATFIV file interpreter was essential to
uncovering the uninitialized variables. We feel it did a thorough
job.
At the outset, the FORTRAN I/O Unit numbers were converted
from constants to integer variables as shown in Figure 3. The
mnemonics RDR, PTR, defined as four byte variables, were used
and their values supplied in a data statement at the beginning
of the program as shown in Figure 4. Not only does this enhance
the readability of the source code, it also means that the unit
numbers for the reader and the printer may be altered by chang-
ing one or two data statements rather than by searching through
the entire source listing for places that the reader and printer
are used. We feel this should be done generally, not just for
programs written for distribution, because it never can be antici-
pated when system programmers may decide to change unit numbers
to upgrade or modify the system.
171
-------
to
602 DU 650
JF(IK£PET)605,6i5,605
605 REAlfl2J902)MPLOTtJl_f J21J3,J«/JS,J6
615
650 CONTINUE
Figure 3a. Sample I/O statements, as received.
U -650- Ls
IF (1^0)605,015,605
605
615™ rtRITE(GRAPMO'L)MPLOT,Jl,J2* J3V J«» J5,J6,JP1
X
650 COf4TINUC
/S'tyre 36. Sample I/O statement, as modified.
-------
INTEGER RDR,PTR>KMC001,FlLSPLrGRAPHO
DOUBLE PRtCISION XNDPtN(10),YO(10)
Figure 4a. Specification of I/O unit mnemonics 4-byte integer variables.
C * »*» -> «> ASSIGN 1/U UNiT NUMDEftfry-CTC, HERE
C *
DATA RDR/l/,PTR/3/
•€-*
•-€-*-
DATA KMCOO1/1O/
-DEFINE-FI L€-i •> -> OON'T FORGET "FILSPL" DEFINITION IN SUBROUTINE FIL3PL
C
Figure 4b. Data definition statements for defined I/O unit variable.
-------
The only significant difficulty we encountered had to do
with the interface to the plotter; this raises an issue that
has been anticipated and for which there appears to be a general
solution when program systems of this nature are being developed.
The actual working code can be written in such a fashion that
I/O interfaces are localized into one subroutine, which clearly
makes all installation-specific references to a particular device
and then, in the rest of the program, all the I/O can appear
symbolically. As an example of this concept, I will describe
how we handled the graphics part of the last four programs of
CIDRS, using modifications to GRAPH as an example.
As shown in Figure 5a, plotter commands available on Southern's
PDP-15 treat the plotter as an I/O device with a unit no. 7.
This way it is quite logical to have statements such as those
shown in Figure 5b to either write to a device or read from the
device (in the latter case it means reading the location of the
pen). However, at the TUCC installation such detailed control
of the plotter is not possible; it can only be accessed through
high-level subroutine calls. Therefore, we went through the
WRITE (7) mode (followed by optional variables, depending on
the value of mode).
Mode
0
1
2
3
4
5
6
7
8
9
10
Pick up the pen
Put down the pen
Move to absolute coordinates. Address with
pen up
Move to absolute coordinates. Address with
pen down
Move to delta coordinates. Address with
pen up
Move to delta coordinates. Address with
pen down
Draw character (see note below)
Set coordinate address
Move to absolute coordinate address
(no pen change)
Move to delta coordinate address (no pen
change )
Set character attributes
Additional
variables
None
None
IX, IY
IX, IY
IX, IY
IX, IY
ICNT, DATA
IX, IY
IX, IY
IX, IY
IXSI2, IYSIZ,
ISIN, ICOS
Figure 5a. Plotter specifications from Reference 2.
wRITE(LOTS)
U™»QNJr^ f Q v * vu %
» f\ ' i */ (k w A ™ A Q f
JYsRND(SY*YB)
_ _
WRl tE~TLOTS) MQDE7IxTfY
Figure 5b. Sample plotter output statement, as received.
174
-------
source code replacing each plotter I/O statement with a call
to a newly written interface subroutine. This type of call
should be valid, independent of any hardware, and the result
is that the program logic now accesses a subroutine, passing
along values through arguments and through common which will
be of use for the plotter. The plotter interface then is written
as being hardware-dependent. In this way all the necessary data
are available to the interface subroutine, both physically and
logically, and all changes relating to the plotter are physically
isolated to the subroutine. Again, our basic approach was to
design the interface subroutine PLOTTR to mimic the originally
used read and write statements. This approach would probably
be simplified if the I/O interface had been written at the outset
as these programs were being developed. To keep things as con-
sistent as possible with the original CIDRS programs and documen-
tation, we elected to mimic the behavior of the PDP-15 plotter
at least in the original code. In Figure 6, it may be seen how
we accomplish this; at the left is the old write statement, which
would move the pen to specified coordinates; to the right is
the equivalent subroutine PLOTTR call. It may be noticed that the
first parameter in the argument list is the "mode" of the corre-
sponding write statement. Thus, Figure 7 shows that when PLOTTR
is entered the first thing that is checked (following some initial
housekeeping) is this mode value, which then determines what
is to be done by the plotter. Once the branch is taken, the sec-
tion of code which directly relates to hardware requirements is
entered. These blocks, as illustrated in Figure 8, will need to
be modified for application at other installations. These blocks
are very installation-specific and one would expect that an
applications programmer would go to this section of the program
and make the necessary changes. Given the presumption used in
making this kind of interface, and knowing the requirements of
his own system then, one would expect the programmer would be
capable of quickly and reliably making the changes necessary to
implement each particular program on his own system. Using this
approach we have exercised all six CIDRS programs and all function
reliably as far as we are able to tell. We have not been able to
extensively test the programs, but, to the best of our knowledge,
the programs do function as they were intended.
GENERAL REMARKS
A data validation section would be useful in CIDRS as a
quality assurance check, and it appears this easily could be
added. For example, in MPPROG it would be appropriate to de-
termine the type of "window" that is reasonable for each input
parameter; this could be done a priori, as well as from the tech-
nical experience of the programmer. For instance, one would
not anticipate that a negative weight would be a valid entry.
Although a weight change less than zero may be physically ob-
served, a negative weight would not be feasible nor would a
175
-------
-J
cr>
c
c
ENTRY
IYcRND(RINC*H)
IF(IX.LE.O) IX«10
TFTiY.LE.O)_IY«10
"JSINS65536*
"JCOSa6553b*
MODY3TO
wRJTEUOTS) MODE*IX,|Y,JSJN,JCOS
IX 5 R NP(SX«XB)
TT=R"ND(SY*YBl ~~ ~— ~~
C SETS CHARACTER/PLOTTED ATTRIBUTES,- RAISE
C & MOVES-TO STARTING -POSITION—-
C- SEf
IX=RND(RINC*W)
IF(IX.LE.O) IX»10
IF(IY.LE.O) IY»10
(COS(TH)f
JCUS=6553<>*
MODE=10
(COS(TH))
;WRITE (LOTS) MODE", 1X71Y *irsf--mm — ~* ^»CALL PLOTTRC2,XB,YBi0,,0,)
RETURN
ENTRY
MODEs 2
J_X Qs R NO ( SX * X )
I Y 0 s H NO 1 3 Y"*YJ
ENTRY CDGRID (I,X,Y»UrM)
MODE=2
IXO«RND(SX*X)
MUDEsl
WRITE(LOTS) MODE
MUDE = 9 _" ~
JMO\)E 8 s8 '" """
LlMifsM'+l
IF (I,EQC2*
_MY2sO
MY1=0
MX 1*5 "
CALL PLOTTR t2VX>T/0.10.)
MODE=1
MOD£s9
MODE8s8.
GO TO 100
IF (I.EQ.a*(I/2))
GO TO ISO
-------
IFCMODE.EQ.14)GU TU 14
' X=XX • " ' " ' "
Y-YY
HSHM
TSTT
DX=HH
OT-I 1
- XS=X*XSCL ~" " """ ""
YS=Y*YSCL
DYScDY*YSCL
IF(ICNT,EQ.UlNRITE<3,i002)ICUT
1002 FORMAT (• »>rflS,'|*r
-------
oo
C THIS BLOCK SERVES CALLS FOR CMODE«0) 00056082
C , . . PICK UP THE PEN 00056083
tfrrO— • -t(MT IttUE -- - — - — -- - - - - - - - - 88-7
C -- . --------- •-*.»*-»--»••»»»•.....».» 00056088
C THIS BLOCK SERVES CALLS FOR (MODE*l) 00056089
C ', . , PUf-PEtt-frQwtt— - - — ~ — ~ -- — - — - - - - — — — t>tf05t»0*0
I--- - CONTINUE - - —• ------------------ ------ ...... - ....... ------- --- ..... — ..... —00056091
PNDM— ' ------------------------- ----------------------------------------- 00056092
OnX3f-Y8yt1»NWO-- — — --- ----- ' — -— — -- r— -- - — .,_™_^OT) 5^4,3
RETURN ©0056094
-«.--. ...»«i»^ ••«.»», 00056Q«?5
_—-_____—_----
€-.--.—. -MOVE- TO- ex, t )> -WITH-PEW-UP
-2 ----- CONTINUE --------
- -- PC-mjP*tP WP— - - -- -—,_ - -._-,
CALL PLOT(XS,YS»IPNUP) 00056100
RETURN 00056!0t
-€»^^CT3-.., -jr» ••JTJ1 «B.»»lg.aam^» ^r^r^-^-^ -."»"m'^ » ,-^-^-^-tHH>5M w
C- THIS BLOCK SERVES CALLS FOR --- ...... — ..... — ---------------- ~~
-------
weight increase of, for example, greater than 50 mg be reason-
able. So, for the stage weights, a data validation section might
include a check that the weight is not negative and the change
is not larger than 50 mg. In terms of data quality and of being
more assured of the functioning of the programs, this type of
approach to screening both the input data and the control param-
eters is being looked into and may become part of a subsequent
version.
Another aspect of data quality that is of concern is the
fact that these programs will be modified to run on different
operating systems. Also, depending on the user's perspective
and needs, he may decide that he wants to modify some of the
basic algorithms to alter the output. In this situation the
program system no longer is CIDRS, but is a local modification
of CIDRS and in any documentation it should be made clear just
what these modifications are. Otherwise, comparing data between
two operating groups will be questionable or impossible simply
because a different physical meaning due to the different mathe-
matics involved may very well have occurred.
In summary, we have obtained CIDRS as distributed by Sou-
thern Research Institute. We have implemented it in an environ-
ment different from that in which it was developed, attempting
to locate and highlight identifiable problems and to design modi-
fications that will be useful to other users installing CIDRS
on their system. CIDRS appears to be a very useful and well-
written set of programs, especially considering that it is the
first generation package.
179
-------
PAPER 11
NUMERICAL SIMULATION STUDIES AND DATA REDUCTION
FOR SIZE CLASSIFYING MEASUREMENT TECHNIQUES
H. FISSAN
C. HELSPER
AEROSOLMESSTECHNIK
GESAMTHOCHSCHULE DUISBURG
ABSTRACT
Nearly all techniques that classify particles to obtain
size distributions show a certain cross-sensitivity. Only a
certain fraction of the particles belonging to one class of par-
ticle size is represented in that class while the rest is found
in the neighboring classes.
Since generation techniques for monodisperse aerosols have
been developed, it is possible to describe this non-ideal instru-
ment behavior by rather straightforward calibration experiments.
The knowledge of this systematic error makes it possible to
simulate the instruments numerically for given size distribu-
tions.
The influence of the non-ideal instrument behavior on the
measured distribution parameters is discussed for a wide range
of possible distributions.
A data reduction program has been tested by simulation
studies for various unimodal and bimodal log-normal distribu-
tions. The influence of random errors in the measured data on
the results of the correction calculations is discussed. Finally
the data reduction program has been applied to several real
aerosols.
INTRODUCTION
For several reasons there is an increasing interest for
the determination of the size distribution of aerosols. For
this purpose various measurement techniques have been developed
which allow classifying the particles according to their size-
dependent behavior into a number of size intervals and deter-
mining a measure for the particle frequency in these intervals.
The size of a particle is normally represented as the equivalent
180
-------
diameter of a sphere which would show the same reaction as the
measured particle. So this equivalent diameter depends very
much on the measurement principle used. The particle frequency
is mostly determined as particle number or particle mass or as
a number or mass concentration.
For all these measurement techniques the classification
procedure as well as the determination of the particle concen-
tration is not free from certain systematic errors which may
cause considerable differences between the real size distribution
and the one determined by the instrument. If this non-ideal
instrument response is known, one can estimate the influence
of these effects and can try to correct the measured data to
obtain a size distribution nearer to the real one.
REAL INSTRUMENT RESPONSE OF SOME CLASSIFYING MEASUREMENT TECH-
NIQUES
This non-ideal instrument behavior, though predicted by
theory in most cases, has often been neglected in the past, be-
cause there was nearly no possibility to describe these effects
exactly. Since monodisperse aerosol standards like the vibrating
orifice generator or the electrostatic classifier are available,
this situation has changed and a lot of work has been done on
the calibration of classifying instruments.
So the electrical aerosol analyzer (EAA) has been calibrated
by Liu et al.1 and its response to monodisperse aerosols has
been represented by a calibration matrix. The cross channel
interference shown by this matrix is mostly due to the mere
statistical distribution of the number of elementary charges
on particles of the same size.
An optical particle counter (type Gertsch HC-15) has been
calibrated by OOP particles in our lab. This instrument has
a very small scattering volume which allows measurement of par-
ticle concentrations up to 105cm~3 with still neglectable coin-
cidence errors. As the cross section of this scattering volume
is smaller than the cross section of the sample stream, a certain
number of particles cross the edge of the volume, thus scattering
less light than a particle in the center of this volume. This
is the reason for a considerable cross channel interference of
this device. Figure Ib shows the instrument response for several
monodisperse aerosols as a cumulative number distribution versus
the particle diameter indicated by the instrument.
The fraction of particles measured too small increases with
increasing particle size because of the increased probability
for those particles to pass partially outside the scattering
volume. Calibration studies in the lower particle size range
are now done by means of an electrostatic classifier. For com-
parison the equivalent results for a Royco 225 particle counter
181
-------
z
o
1.01
S3 0.8
cc
t-
Q
cc 0.6
LU
m
D
ui 0.4
= 0.2-
O
ROYCC 225/241
10
PARTICLE DIAMETER Dp,
1.0
m 0.8-1
E
i-
cc
U
00
0.6-
z 0.4 H
LU
0.2-
O
HC IS
B)
23 5
PARTICLE DIAMETER.Dp, /urn
10
Figure 1. Calibration results of optical particle counters.
are shown in Figure la. The scattering volume of this instrument
is limited by the sample stream so that all particles pass it
with the whole cross-section. Therefore the instrument behavior
is nearly ideal.
182
-------
Similar calibration experiments have been done for a six-
stage Andersen Stack Sampler (A.S.S.)2 with liquid oleic acid
particles to determine the collection efficiencies of the stage
as well as the wall losses. In Figure 2 the percentage of total
wall loss is plotted against the aerodynamic particle diameter.
Results of Gushing et al.3 for an eight stage A.S.S. and solid
particles are given for comparison. Besides the fact that the
number of stages was different in the two experiments, the dif-
ferences between the two curves might be caused by the following
reasons:
1. flow, particle density, and temperature were not the
same in the two experiments,
2. the adhesion of solid and liquid particles on walls
might be different,
3. in Gushing's experiments the impactor was mounted hori-
zontally, while in ours it was mounted vertically.
CO
O
80
70
60
50
40
30
20
10
0
IMP ACTOR
V A.S.S. 6
D A.S.S. 8
PARTICLES FLOW RATE
LIQUID
SOLID
22.6 l/min (20°C, 1 BAR)
14.2 l/min (22°C, 1 BAR)
23 5 10 15
AERODYNAMIC PARTICLE DIAMETER, DAE, jum
20
Figure 2. Total wall losses of Andersen Stack Sampler (Gushing, et al.3).
183
-------
The collection efficiencies determined in the two experiments
are compared in Figure 3. Due to bounce off and blow off the
collection efficiencies for solid particles in no case reach
a value of 100%. For liquid particles this extreme shape of
the collection efficiency curve is only found for the first
stage, where the wall losses cause this instrument response.
INFLUENCE OF NON-IDEAL INSTRUMENT RESPONSE ON THE DETERMINATION
OF SIZE DISTRIBUTIONS
If one has determined the instrument's response for a set
of monodisperse aerosols, it is possible to establish a response
matrix to describe the instrument's overall behavior. This makes
it possible to simulate the instrument numerically and to predict
its response to any given size distribution for an aerosol similar
to the one used in the calibration procedure. The comparison
of the instrument's output with the input distribution allows
estimation of the errors caused by its non-ideal behavior.
Results of a simulation of the EAA for two log-normal dis-
tributions are shown in Figure 4. The cumulative number distri-
bution is plotted against the particle diameter. The geometric
standard deviation for both input distributions which are repre-
sented by the solid lines is Og = 2.0; the number mean diam-
eters are 0.03 ym and 0.3 ym, respectively. In both cases the
EAA tends to shift the mean diameter of the output distributions
(given by the circles and dashed lines) to smaller values while
the shape of the distribution is not very much affected. The
influence on the number mean diameter (NMD) seems to be dependent
on the input NMD.
This dependency is shown in more detail in Figure 5. The
ratio of the NMD calculated from the simulated output data to
the actual input NMD is plotted as a function of the input NMD
for three geometric standard deviations Og. For an ideal in-
strument this ratio should be one, at least for distributions
in the middle of the covered size range, while near the upper
and lower limit of this range, there are deviations due to the
fact that the instrument covers only a part of the distribution.
But the real instrument shows output NMD's which are by 20% too
small even in the middle of its range. For NMD's smaller than
0.1 ym there are points where the effects of limited size range
and non-ideal behavior compensate. Around these points the in-
strument will determine a correct output NMD.
If one calculates under the assumption of spherical par-
ticles a volume distribution out of the measured number distri-
bution there will be deviations between the actual input volume
mean diameter (VMD) and the calculated output VMD, too. Figure
6 shows the ratio of calculated to actual VMD, plotted against
the input NMD. The deviations in this case are even more ob-
vious, especially for large standard deviations, which is mainly
184
-------
COLLECTION EFFICIENCY,
§
§
00
U1
;n
I
.
§
3!
fV
S'
O
-*>
^
I
1
I
p
(JI
3D
o
D
I a
m
a>
D
oo
O
CO
COLLECTION EFFICIENCY,
s
§
30
O
0
30
H
O
g
>
m
30
-------
98
0.01
0.1 1.0
PARTICLE DIAMETER, jum
Figure 4. Calculated electrical aerosol analyzer response for given log-normal
size distribution.
due to the limited size range. As the EAA sensitivity to number
concentration depends strongly on particle size, the number con-
centration determined with this instrument is also affected by
its cross-channel interference. Figure 7 shows the ratio of
the total calculated output number concentration to the actual
input concentration. For distributions in the middle of its
size range the instrument tends to give a number concentration
reading which is about 10% too high, while near the limits of
the size range too small readings are caused by the fact that
only a fraction of the particles is counted.
186
-------
Ill
o
o
0.01
1.0
NUMBER MEAN DIAMETER NMD, j
Figure 5. Ratio of calculated and actual number mean diameters for an
electrical aerosol analyzer.
Q
5
D
_l
O
O
0.01
NUMBER MEAN DIAMETER NMD, jum
Figure 6. Ratio of calculated and actual volume mean diameters for an
electrical aerosol analyzer.
187
-------
= 1.4
S\s
Z
O
P
O
LLJ
CO
D
Z
ui
U
Ui
u
oc
LU
m
D
0.01
1.0
NUMBER MEAN DIAMETER NMD, Mm
Figure 7. Ratio of calculated and actual number concentration for an electrical
aerosol analyzer.
Similar calculations have been done for a six-stage A.S.S.
Figure 8 shows the results for three log-normal distributions
(<7g = 3, MMD = 1.5 and 10 pm) as a cumulative mass distribution
plot. For these calculations the collection efficiencies for
liquid particles have been used. One can see that the deviation
between the input distributions (solid lines) and the simulated
measured data is nearly neglectable for distributions with small
mass median diameter (MMD). It tends to increase with increasing
MMD, mainly due to increasing wall losses.
These wall losses have been simulated for solid and liquid
particles for a set of input distributions. Figure 9 shows these
wall losses as a function of MMD and ag. In both cases the wall
losses reach values of more than 30% of the total mass for large
MMD's. For solid particles they do not very much depend on the
geometric standard deviation of the input distribution.
The ratio of calculated output MMD to actual input MMD as
a function of the input distribution parameters for liquid par-
ticles is shown in Figure 10. As the impactor does not have
a lower size limit because of its backup filter, its behavior
is nearly ideal for small MMD's. For larger MMD's and large
erg's the output MMD is much smaller than the one for the input
distribution. Besides the effect of the upper size limit, the
increasing wall losses are the reason for this behavior.
188
-------
CUMULATIVE MASS DISTRIBUTION, %
m
30
O
O
P
ui
oo
VD
-
I
cf
Ol
>
m -»
H o
m
30
Q.
Co'
5?
Cr
C
***•
8'
s
en
I I I I
Is)
O
S g
1—r
CO
vt
to
CO
I I I I
to
I I I
I I
I I I I
-------
30
20
10
ta
01
_J
_]
s
30
20
10
COLLECTION EFFICIENCY FOR LIQUID PARTICLES
(V=22.6 l/min)
COLLECTION EFFICIENCY FOR SOLID PARTICLES
(V=14.2 l/min)
-------
0.7-
0.6 -
10
MASS MEDIAN DIAMETER MMD.
Figure 10. Ratio of calculated and actual mass median diameters for A.S.S. (collection
efficiencies for liquid particles; V = 22.6 l/min).
DATA REDUCTION TAKING INTO ACCOUNT NON-IDEAL INSTRUMENT BEHAVIOR
These instrument simulation studies show that there are
cases in which the errors caused by the instrument response can-
not be neglected.
A data reduction procedure which would allow correction
of the measured data should fulfill the following requirements:
1. It should allow correction for the instrument's sys-
tematic error even if the response matrix of the in-
strument is highly unstable (small errors in the mea-
sured data will give large changes in the solution).
2. It should reduce the effect of random errors in the
measured data by use of physical constraints.
3. It should reduce the amount of data necessary to de-
scribe the particle size distribution.
4. It should allow a certain extrapolation of the results
over the size range covered by the instrument. This
would allow a comparison of different measurement tech-
niques with only small overlapping size ranges.
191
-------
To obtain these requirements, it seems to be reasonable
to make the assumption of a certain type of size distribution.
It has been shown1* that by superposition of two or three log-
normal distributions most aerosol size distributions can be de-
scribed. For such a distribution a set of parameters has to
be determined, for which the deviation between the simulated
output data of the instrument and the actual measured data be-
comes a minimum.
A simplex minimization procedure (suggested to us by B.Y.H.
Liu and A. Kapadia) for the EAA, has been chosen to optimize
the distribution parameters. A modified chi-square with an ad-
ditional weighting function was chosen as an objective function
to characterize the goodness of fit between the simulated and
the measured data. Starting parameters for the minimization
procedure are estimated from the measured data. As the optimi-
zation algorithm does not necessarily detect the absolute mini-
mum, several restarts with random variations of the starting
parameters can be done.
INVESTIGATION OF THE EFFICIENCY OF THE DATA REDUCTION PROCEDURE
A lot of calculations were done to test if this data reduc-
tion procedure works satisfactorily. In a first step the program
was checked with input data which were simulated from unimodal
and bimodal starting parameters. The summarized results of these
studies are shown in Table 1.
In test no. 1 the program used fixed starting parameters,
which were not estimated from the input data. In the case of
the unimodal distribution the parameters of the input distri-
bution were recovered for all considered cases (12 sets of param-
eters) . The parameters of the bimodal distribution were recovered
only in 41% of a total of 243 sets of distribution parameters.
The calculation of estimates of the starting parameters
from the simulated input data raised the success rate for bimodal
distribution to 59%. In test no. 3 up to ten restarts with
random variation of the estimated starting parameters in a range
of ± 30% were allowed, if the goodness of fit indicated by the
absolute value of the objective function was not satisfactory.
This finally gave a success rate of 94% which seems to be the
best that could be achieved by this method.
In some few of the successful cases the actual parameters
of the input distribution were not recovered in spite of the
fact that the value of the objective function indicated a very
good fit. The recovered distribution was compared to the input
distribution in a plot and agreed nearly exactly, which means
that quite different parameters can describe nearly the same
distribution. This case only occurred when the two distribution
modes lay close together.
192
-------
To get results which were closer to reality, in a next step
the simulated input data were slightly varied to simulate random
errors: 12 unimodal and 25 bimodal distributions with ten dif-
ferent sets of errors applied to each were investigated. In
test no. 4 the errors were equally distributed between 0 and
10%, and in test no. 5 between 0 and 15%. The success rate in
both cases was 100% for the unimodal and around 85% for the
bimodal distributions. To test the influence of errors on the
response matrix elements, which could be due to the limited
accuracy of the calibration experiments, calculations were done
with random errors (0 - 5%) in the matrix elements. The input
data were free from errors. The success rate for the unimodal
and bimodal distributions was 100%.
As a conclusion one can say that the data reduction pro-
cedure works very well in the case of unimodal distributions.
For bimodal distributions several restarts with varied starting
parameters are absolutely necessary to get a considerable rate
of success. Even then there are cases in which no solution is
found. With random errors in the input data the success rate
decreases slightly.
TABLE 1. RESULTS OF TEST CALCULATIONS FOR THE EAA
DATA REDUCTION PROGRAM
Success rate
Test no. Test conditions Unimodal Bimodal
1 Simulated data, fixed starting
parameters 100% 41%
2 Simulated data, starting
parameters estimated from
the input data 100% 59%
3 Like test no. 2, up to ten
restarts with random varia-
tions of the estimated
starting parameters 100% 94%
4 Like test no. 3, with random
errors in the input data;
errors equally distributed
between 0 and 10% 100% 83%
5 Like test no. 4, errors equally
distributed between 0 and 15% 100% 85%
6 Like test no. 3, random errors
in the response matrix
elements (0 - 5%) 100% 100%
193
-------
When applying this data reduction method to real data, the
random errors in these data will make success more difficult;
the instrument behavior as determined in the calibration experi-
ment may not be the same for particles of a shape and material
different from the calibration aerosol.
To eliminate these influences, a first application check
was done for a polydisperse DOP aerosol produced by an atomizer,
Figure 11 shows the results of this experiment. The particle
107 -
co
o
a
Q
O
o
106-
<
cc
111
o
o
o
cc
111
CO
111
_l
O
I-
cc
<
a.
104
CHISQ = 0.014
— CORRECTED NUMBER DISTRIBUTION
P MEASURED DATA
• SIMULATED DATA
0.01
0.1
PARTICLE DIAMETER Dp,
Figure 11. Atomizer aerosol (DOP).
194
-------
number concentration ACN/Alog Dp is plotted against the particle
diameter Dp. The measured data are represented by squares.
The corrected unimodal number distribution as determined by the
data reduction program is given by the solid line. The simulated
output data for this distribution, which in the case of an ab-
solute fit had to be identical with the measured data, are given
by circles. The deviations between measured and simulated data
are in the range of the instrument's accuracy, which means that
the corrected number distribution represents one possible solu-
tion. The value of the objective function is CHISQ = 0.014,
which indicates a good fit, too.
APPLICATIONS OF THE DATA REDUCTION PROCEDURE TO REAL AEROSOLS
As mentioned above, real aerosols may differ from the cali-
bration aerosols, as far as particle shape, particle material,
and state of the gas are concerned. For any one of these reasons
the response matrix of an instrument may not be valid for a cer-
tain aerosol which will cause a failure of the data reduction
procedure.
In the following, several examples of applying this pro-
cedure to data of different aerosols are given.
The EAA has been used to determine the particle number dis-
tribution of the aerosol emitted by a spark ignition engine in
connection with studies to characterize the content of polyaro-
matic hydrocarbons (PAH) in the particulate matter emitted by
the engine. Figure 12 shows an example of a size distribution
of this exhaust aerosol. The solid line gives the bimodal cor-
rected number distribution and the squares represent the measured
data, while the circles give the simulated data. As in the
previous example, the deviations between the measured and simu-
lated data are in the range of the instrument's accuracy, which
leads to a rather small value of the objective function.
In another study the EAA has been used for the determina-
tion of the size distribution of test fire aerosols. These test
fires are used to test the reaction of automatic fire detectors.
In Figure 13 the particle number distribution of a smoldering
wood fire aerosol is given. The fit between the measured and
simulated data is very good; the value of the objective function
is very low.
The next example, which is shown in Figure 14, represents
the size distribution of a polyurethane fire aerosol. In this
case the fit is very bad. The geometric standard deviation of
the second mode has a value of Oq2 ~ !•!/ which was the lower
limit for the optimization procedure. Even with this narrow
distribution the simulated data are more broadly distributed
than the measured data, which means that for a better fit the
input distribution had to be even narrower. As it seems very
195
-------
unlikely that a fire will produce a monodisperse aerosol, this
seems to be a case in which the instrument's calibration is not
valid. This, however, makes it impossible to find an appropriate
solution with a data reduction procedure based on this calibra-
tion.
108-
co
o
a
O
o
<
10
,7-
01
o
o
o
cc
Ul
CO
§
o
cc
106-
105
a
CHISQ = 0.021
— CORRECTED NUMBER DISTRIBUTION
D MEASURED DATA
• SIMULATED DATA
0.01
—i—
0.1
PARTICLE DIAMETER Dp, fim
Figure 12. Spark ignition engine exhaust aerosol n = 3000 min~1, X = 0.92.
196
-------
In Germany, a certain type of SiO2 aerosol is used in test-
ing instrumentation in a dust tunnel. The particle mass distri-
bution of this aerosol has been determined by Geipel5 with an
Andersen Stack Sampler. In Figure 15 his results are shown.
In a first attempt we tried to correct his data assuming a log-
normal size distribution and using the collection efficiencies
for solid particles. The resulting log-normal size distribution
10? -
n
o
a
a
o
<
106-
LLJ
O
O
O
EC
111
CQ
a
o
P
OC
<
O-
105-
104
•
D
CHISQ = 0.0076
— CORRECTED NUMBER DISTRIBUTION
D MEASURED DATA
• SIMULATED DATA
0.01 0.1
PARTICLE DIAMETER D
Figure 13. Smoldering wood fire aerosol.
197
-------
shows rather good agreement with the distribution determined
with sedimentation analysis. The calculated wall losses (36%)
are also comparable with the ones measured by Geipel (38%).
On the other hand the value of the objective function is rather
10?-
106 -
a
a
o
_i
<
o
2
O
CC
I-
LU
u
2
O
o
cc
LLJ
CO
D
a, 105
104
a
CHISQ = 0.37
CORRECTED NUMBER
DISTRIBUTION
D MEASURED DATA
• SIMULATED DATA
0.01
0.1
PARTICLE DIAMETER D
p/
Figure 14. Polyurethane fire aerosol.
198
-------
large and also the differences between measured and simulated
data. Here again it hasn't been verified that one can use the
collection efficiencies of the calibration experiment for this
real aerosol.
10
,3 -
CO
•
Q.
Q
O
-------
SUMMARY AND CONCLUSIONS
Almost all classifying instruments show a certain cross
sensitivity, which can be determined by applying a monodisperse
aerosol. Knowing the behavior of the instruments one can simu-
late the instrument numerically. These simulation studies, here
shown for an Electrical Aerosol Analyzer and an Andersen Stack
Sampler, allow us to judge the importance of including real
behavior in the data reduction. A data reduction program based
on the simplex minimization procedure has been carefully checked.
It has been found that in almost all cases, even if errors are
involved, a solution is found. In applying it to real aerosols
this data reduction procedure finds its main limitation in the
lack of knowledge about the instrument behavior for the real
aerosol, which may be quite different from the aerosol used in
the calibration experiment.
REFERENCES
1. Liu, B.Y.H., and D.Y.H. Pui. On the Performance of the
Electrical Aerosol Analyzer. J. Aerosol Sci. 6:249-264,
1975.
2. Franzen, H., H.J. Fissan, and U. Urban. Eichung eines
Andersen-Stack-Samplers unter Verwendung des Berglund-Liu-
Generators. [Calibration of an Andersen Stack Sampler with
the Use of a Berglund-Liu Generator.] Staub Reinhalt. Luft
38:436, 1978.
3. Gushing, K.M., G.E. Lacey, J.D. McCain, and W.B. Smith.
Particulate Sizing Devices for Control Device Evaluation:
Cascade Impactor Calibrations. EPA-600/2-76-280, U.S. En-
vironmental Protection Agency, Research Triangle Park, NC,
1976. 94 pp.
4. Whitby, K.T. Modelling of Multimodal Aerosol Size Distri-
butions. Presented at the Annual Meeting, Gesellschaft
fur Aerosolforschung, Bad Soden, Germany, October, 1974.
5. Geipel, W., and R. Wiedemann. Erprobung von Emissionsmessver-
fahren zur Feststellung von Korngrossenfraktionen. [Tests
of Methods for Measuring Emissions to Determine Particle
Size.] Report No. 7/79. Lehrstuhl fur Thermische Kraftan-
lagen, Technische Universitat Munchen.
200
-------
PAPER 12
NON-IDEAL BEHAVIOR IN CASCADE IMPACTORS
JOSEPH D. McCAIN
JAMES E. McCORMACK *
SOUTHERN RESEARCH INSTITUTE
INTRODUCTION
Cascade impactors have become commonly used measurement
devices for the determination of size distributions of parti-
culate emissions from industrial sources. Data obtained by means
of impactors are used to characterize emissions from sources,
to determine the performance of particulate control devices,
and in selecting and designing control devices for specific
sources.
Data provided by impactors are of relatively low resolution
and do not permit the exact reconstruction of the size distri-
bution of the aerosol being sampled, even over the limited range
of sizes nominally covered by most impactors (approximately 0.5
to 15 jam). However, little has been done to estimate the magni-
tude of the uncertainties, or errors, which are inherent in the
method insofar as they relate to industrial source emission mea-
surements and determinations of fractional collection efficiencies
of control devices. The study described here was one with the
specific goals of estimating the effects of two non-ideal operat-
ing characteristics of impactors on the data obtained with them.
These two non-idealities are (1) the lack of step function stage
collection characteristics and (2) particle bounce. Several
authors1'2'3 have proposed various deconvolution procedures which,
when applied to impactor data, would, to a large degree, correct
for the effect of the finite slopes of the stage collection ef-
ficiency curves. However, little use has been made of these
procedures, primarily because noise in the data frequently results
in oscillatory solutions with large negative values. In any
case, little quantitative information regarding the magnitude
*Now at the University of Minnesota
201
-------
of the errors introduced by the lack of sharp size cuts in im-
pactors commonly used for stack sampling has been published.
The magnitude of errors introduced by particle bounce has not
previously been quantified at all although the existence of such
errors has been described in the literature. **'5'6'7
TECHNICAL PROCEDURES
The approach used in this study was the development of a
computer model of cascade impactor performance. The model was
based on actual impactor performance as measured in a calibra-
tion study of commercially available cascade impactors for stack
sampling. A total of four simulation models were used for both
a Brink impactor in a commonly used modified configuration for
stack sampling and an Andersen Mark III stack sampler. The use
of glass fiber collection substrates was assumed for both im-
pactors. Both greased substrates and glass fiber substrates
are commonly used for sampling at temperatures below 150°C (300°F)
but no satisfactory greases have been found for use at tempera-
tures over 150°C. Therefore, glass fiber substrates must usually
be used for collection substrates at elevated temperatures.
The first model for each impactor was one having ideal col-
lection characteristics, i.e., step functions from 0% to 100%
collection at the stage D50's. (The stage D50 is that particle
diameter at which the stage has a collection efficiency of 50%.
The D50 is generally used as the characteristic cut off diameter
for particles collected by the stage.) This model was used as
a performance standard against which the remaining three models
could be compared and also provided a basis for checking the
program.
The assumed operating conditions and resulting cut sizes
(Dso's) of the two impactors modeled in the study are given in
Table 1. The models of the Brink impactor included a cyclone
precollector which was assumed to have the same performance
characteristics in all three of the simulations other than that
of the "IDEAL Brink." The cyclone performance was based on
calibration data for a cyclone in common use with the Brink im-
pactor in a modified configuration for stack sampling. This
cyclone had a collection efficiency of 100% for particles larger
than about 20 ym.
The second model for each impactor utilized the actual cali-
bration data for each stage. In this model the stage collection
efficiencies increased monotonically with increasing Stokes1
numbers (increasing particle size) to a maximum value of about
90% to 95%. The efficiencies then decreased for larger Stokes1
numbers to a value of 35% to 40% and remained constant there-
after. A composite of the calibration data for stages two (2)
through seven (7) of the Andersen impactor, which illustrates
202
-------
TABLE 1. SIMULATION CONDITIONS OF THE MODELED IMPACTOR
PERFORMANCE
Temperature, °C (°F)
Gas composition
Particle density, g/cm3
Flow rate, alpm (acfm)
Barometric pressure, mm Hg
Stage/D50(ym)
Brink
177 (350)
Standard air
2.27
1.13 (0.040)
749
Andersen
177 (350)
Standard air
2.27
17.0 (0.600)
749
1
2
3
4
5
6
7
8
9.
6.
3.
1.
1.
0.
0.
0.
60a
18b
39
85
23
905
589
198b
7.84
7.40
4.44
2.87
1.58
0.855
0.449
0.200
a Stage 1 is a
Stages 2 and
cyclone precollector .
8 are part of the modifications
to the impactor.
the behavior described above, is shown in Figure 1. Data for
stages 1 and 8 were offset from the tight grouping of the data
for the remaining stages and hence were omitted in Figure 1 for
purposes of clarity in illustrating the behavior trends of the
stage efficiency curves. The data for the Brink impactor ex-
hibited similar trends. This model, Model 2, is called the
"Normal Bounce" model.
The third model was identical to the second except that
the rollover and decline in efficiency for larger Stokes1 num-
bers was ignored. Instead, the efficiencies were assumed to
smoothly increase to 100% and remain at that value for increas-
ingly larger Stokes1 numbers. This is called the "No Bounce"
model. The fourth model was also identical to Model 2 with the
exception that the collection efficiencies were assumed to drop
rapidly to a value of 2% for Stokes1 numbers larger than that
at which the collection efficiency reached a peak in the cali-
bration data. This model was termed the "Extreme Bounce" model.
The use of the same basic collection efficiency curves for
the "No Bounce", "Normal Bounce", and "Extreme Bounce" models
for particle sizes smaller than those for which the collection
efficiencies were maximal in the calibration data is probably
a realistic representation of the actual performance of the im-
pactors in collecting various types of particles. Rao5 found
that impactor collection characteristics for dry solid particles
203
-------
and oil particles were virtually identical when glass fiber sub-
strates were used for Stokes1 numbers smaller than those at which
the peak efficiency was reached for the dry solid particles.
Beyond this point he found that oil particles were collected
with efficiencies which increased to 100% with increasing Stokes'
number while the efficiencies declined for the dry particles
as a result of bounce. Figure 2 shows an example of the four
modeled collection efficiency characteristics of one stage of
the Andersen impactor.
RESULTS AND DISCUSSION
The performance of each model of the two impactors was
evaluated for aerosols having log-normal size distributions with
mass median diameters (MMD) of 1.5, 2.6, 4.5, 7.8, 13.5, and
27 micrometers. Geometric standard deviations, ag, of 2, 3,
and 4 were used at each particle size. Because or the volume
of the data generated only representative results are presented
in this paper.
98
90
70
s?
o
z
Ul
o 50
LU
Z
o
o
LU
O
O
ai
O
CO
30
10
0.5
0.1
0.01
i
0.04 0.1 1.0
SQUARE ROOT OF STOKES NUMBER (DIMENSIONLESS)
10.0
4181-141
Figure 1. Composite of calibration data for the Andersen impactor, stages 2 through 7,
204
-------
Figures 3 through 6 show typical results for the two im-
pactors in a cumulative percentage presentation. It is evident
from all four of these figures that particle bounce severely
distorts the size distributions, especially for aerosols having
large mass median diameters. Figures 3 and 4 show the results
of the simulations for the same impactor and size distributions,
the difference between the two being the omission of the back-
up filter catch in presenting the results in Figure 4. Comparison
of Figures 3 and 4 indicates that omitting the back-up filter
in calculating the cumulative percentages greatly reduces the
distortion resulting from bounce. Comparison of Figures 3 and
5 shows that increasing the width of the input size distribution
(increasing ag) reduces the distortion caused by bounce although
the distortion remains appreciable for the extreme bounce models
at large HMD's.
100 —
5 10 20
PARTICLE DIAMETER, ^m
100
4181-142
Figure 2. An illustration of the four modeled stage collection efficiency
curves of a typical stage of the Andersen impactor. Model 1 is
the ideal behavior model, model 2 is the normal bounce modelt
model 3 is the no bounce model, and model 4 is the extreme
bounce model.
205
-------
Ill
N
V)
X
cc
LU
_l
_l
<
GO
99.8
98
90
70
50
30
z
ui
O
cc
£ 10J-
LU
D
O
2 _
0.5 -
0.1 -
0.01
0.1
• NO BOUNCE
A NORMAL BOUNCE
• EXTREME BOUNCE
1.0
PARTICLE DIAMETER,
10.0
4181-143
Figure 3. Recovered particle size distributions on a cumulative percentage basis from
the Brink impactor models for og = 2.0 and mass median diameters of 1.5,
4.5, 13.5, and 27 urn. The heavy fines represent the input distributions.
206
-------
99.8
N
cc
ai
-i
00
01
u
cc
UJ
Q.
O
98
90
70
50
30
10
0.5
0.1
0.01
0.1
• NO BOUNCE
A NORMAL BOUNCE
• EXTREME BOUNCE
1.0
PARTICLE DIAMETER, fjm
10.0
4181-144
Figure 4. Recovered particle size distributions on a cumulative percentage basis from the
Brink impactor models shown in Figure 3 with backup fitter catches omitted
from the analysis. The heavy lines represent the input distributions.
207
-------
99.8
98 —
LU
N
35
2
cc
Ul
GO
Ul
Ul
o
cc
Ul
Q.
Ul
o
• NO BOUNCE
A NORMAL BOUNCE
• EXTREME BOUNCE
0.01
1.0
PARTICLE DIAMETER,
10.0
4181-145
Figure 5. Recovered particle size distributions on a cumulative percentage basis from the
Brink impactor models for Og = 3.0 and mass median diameters of 1.5, 4.5,
13.5, and 27 fj.m (backup filter included in the analysis). The heavy lines
represent the input distributions.
208
-------
99.8
N
CO
Z
X
cc
98
90
70
50
<
CO
UJ
C3
< 30
Ul
o
DC
UJ
o.
ui 10
D
5
D ?
O z
0.5
0.1
0.01
« NO BOUNCE
A NORMAL BOUNCE
• EXTREME BOUNCE
0.1
PARTICLE DIAMETER, jum
10.0
4181-146
Figure 6. Recovered particle size distributions on a cumulative percentage basis for the
Andersen impactor for OQ = 2.0 and mass median diameters of 1.5, 4.5, and
13.5 yjn. The backup filter was excluded from analysis in the results shown.
The heavy lines represent the input distributions.
209
-------
Figure 6 shows the results from the Andersen models corre-
sponding to three of the four cases for the Brink model shown
in Figure 4. Note that the deviations from the input distri-
bution resulting from bounce are more severe in the Andersen
results than in the Brink. This difference in the severity of
the distortions apparently results from the cyclone precollector
on the Brink which removes most of the large particles which
are responsible for raising the apparent percentages of fines
in the recovered size distributions.
It should also be noted that the relative errors in mass
median diameters become increasingly large as the MMD of the
input aerosol decreases. The recovered MMD's were found to be
systematically large for test aerosol MMD's below 10 ym. The
values of ag were also systematically high with larger relative
errors at the lower values of og, as would be expected because
of the low resolution afforded 5y impactors.
In many cases (e.g., control device fractional collection
efficiency studies) the slope of the size distribution curve,
expressed in mass concentration units, is the quantity of greatest
interest. The most common manner of presentation of this slope
is of the form dm/dlogD versus diameter (units of mass concentra-
tion) . The quantity dm/dlogD is often approximated directly
from the impactor data, stage by stage, as Am^/logD^, where
Am^ = mass concentration of particles retained by the ith
stage
and AlogD. = log —77;—: .
1 (DsoJi
The particle diameter is then taken to be the geometric mean
of (DSQ)..^ and (Dgo).^ or
Dg = (Dso^ x (D50)i_1
Figures 7 and 8 illustrate recovered size distributions
presented in such a manner, together with the input distribu-
tions, for representative sets of Brink and Andersen results.
The results for the "extreme bounce" case are not shown in Fig-
ures 1 and 8 but the values in those cases generally fall between
the "no bounce" and "normal bounce" cases except for the back-
up filters, for which the values were much higher in the "extreme
bounce" case than in the other two cases. Except for the finest
size fractions, represented by the back-up filter catches, and
the fine fraction tails of the low ag distributions, the agree-
ment between the recovered values of (Am/AlogD)^ generally lie
reasonably close to the input distributions. However, errors
of up to ±35% are not infrequent.
210
-------
101
10°
C
10"
g
<
s
10
,2
10"'
10'4
o NO BOUNCE
A NORMAL BOUNCE
- MMD = 4.!
ag=2
10-n
O NO BOUNCE
A NORMAL BOUNCE
10'° 101 10'1 10"°
GEOMETRIC MEAN DIAMETER, micrometers
4181-147
Figure 7. Recovered particle size distributions on a differential basis from the Brink impactor
models for mass median diameters of 4.5 and 27 JU/T? and Og's of 2 and 3.
The heavy curves represent the input distributions.
211
-------
•II1
O NO BOUNCE
A NORMAL BOUNCE
i
-------
Tables 2 and 3 show the errors, expressed as percentages,
in the recovered values of (Am/AlogD)i for several cases for
each of the two impactors. (For the purpose of calculating log
D and D for the filter catches, it was assumed that the diameter
range covered by the filter was (D50)8 down to %(D50)8.
Although no results for ag = 4 have been shown, the agree-
ment between the recovered size distributions and the input dis-
tributions was progressively better as ag increased and was quite
good in all cases for ag = 4 with the exception of back-up filter
catches when bounce was present.
Table 4 shows the percentages of cases in which the recovered
value (Am/AlogD)£ lay within factors of 1.2, 1.5, and 2 of the
true value. From these results it appears that the concentra-
tions of fine particles as measured with impactors can seldom
be taken to be known better than to within a factor much smaller
than 1.5 unless the particles are known to be adhesive or an
effective adhesive coating can be applied to the substrates.
TABLE 2, PERCENT ERRORS IN Aitl/AlogD, ANDERSEN IMPACTOR
No Bounce
MMD
Stage/Error
F
8
7
6
5
4
3
2
1.5
144
9
-8
-18
-14
6
4.5 13.5
56
13
-19 18
-26 -20
-19 -30
-23 -28
Normal Bounce
1.5
22
-14
-11
-10
-12
-7
10
9
4.5
38
-5
-4
-8
-17
-21
-15
-22
13.5
4
-24
-25
0
-24
F
8
7
6
5
4
3
568
22
-6
-20
16
5
121000
920
111
9
-18
-27
-20
293
39
-17
-30
44
-12
-11
-12
-12
-7
9
173
8
1
-11
-16
-21
-16
1720
103
37
4
-11
-24
-25
NOTE: Values are omitted for stages for which the collected mass
would be too small to be detected in field sampling programs,
213
-------
TABLE 3. PERCENT ERRORS IN Aro/AlogD, BRINK IMPACTOR
No Bounce
HMD
Stage/Error
F
6
5
4
3
2
1
1.5 4.5 13.5
220
-6
-43
-19
-15
-1
88
220
-35
3
-3
-10
-14
27
84
-8
-15
33
7
1.5
24
-37
-32
-9
-4
4
34
4.5 13.5
53
-21
-39
-12
~9
-20
-3
13
-40
-5
-3
-23
-14
27
-37
4
8
-22
-14
Bounce
F
6
5
4
3
2
1
550
-10
-40
-20
-16
-1
86
238
-8
10
-4
-28
-15
107
7 100
-18 3
41
-39
-29
-10
-4
4
33
123
-24
-33
-12
-10
-18
-4
12
-26
0
-3
-20
-16
-10
17
10
-16
-17
NOTE: Values are omitted for stages for which the collected
mass would be too small to be detected in field sampling
programs.
TABLE 4. PERCENTAGE OF TRIAL CASES IN WHICH RECOVERED
VALUE OF (Am/AlogD) IS WITHIN THE INDICATED
FACTOR OF THE TRUE VALUE
Andersen
Brink
Factor
Stage/Percent
of cases
1.2
1.5
2.0
Factor
Stage/Percent
of cases
1.2 1.5 2.0
F
8
7
6
5
4
3
2
0
57
65
65
75
30
40
(30)
31
71
76
80
100
100
90
(90)
38
79
82
90
100
100
90
(90)
F
6
5
4
3
2
1
0
35
11
80
74
42
71
33
59
47
95
91
96
72
56
82
100
100
91
100
100
Andersen table covers all cases
with HMD = 1.5, 2.6, 4.5, 7.8,
13.5 and ag = 2, 3
for both normal bounce and no
bounce
Brink table covers all cases
with HMD = 1.5, 2.6, 4.5, 7.8,
13.5, 27 and ag =2, 3
for both normal bounce and no
bounce
214
-------
There is some evidence,8'9 although it is not conclusive,
that the use of adhesive coatings (greases) on the substrates
may become ineffective as the particulate deposits build up under
the impactor jets. This would result in the same type of errors
due to particle bounce resulting in back-up filter contamination
by oversize particles with greased substrates as has been shown
to occur with glass fiber substrates.
CONCLUSIONS AND RECOMMENDATIONS
From the evidence presented here, it is suggested that back-
up filter catches generally should be omitted from data presenta-
tion when dry, non-sticky particulates are sampled. Exceptions
should be made only if the MMD is smaller than about 2.5 ym.
In addition it is suggested that cyclone precollectors having
D50's somewhat larger than the first impaction stage D50 be used
whenever a non-sticky particulate is sampled. The use of such
cyclones tends to greatly reduce errors due to particle bounce.
ACKNOWLEDGEMENT
The work reported here was done under support from the In-
dustrial Environmental Research Laboratory of the U.S. Environ-
mental Protection Agency; Contract No. 68-02-2131, D.B. Harris,
Project Officer.
REFERENCES
1. Cooper, Douglas w., and John W. Davis. Cascade Impactors
for Aerosols: Improved Data Analysis. Am. Ind. Hyg. Assoc,
J. 33(2):79-89, 1972.
2. Cooper, Douglas W., and Lloyd A. Spielman. Data Inversion
Using Nonlinear Programming with Physical Constraints:
Aerosol Size Distribution Measurement by Impactors. Atmos.
Environ. 10(9):723-729, 1976.
3. Picknett, R.G. A New Method of Determining Aerosol Size
Distributions from Multistage Sampler Data. J. Aerosol
Sci. 3:185-198, 1972.
4. McCain, J.D., K.M. Gushing, and W.B. Smith. Methods for
Determining Particulate Mass and Size Properties: Labora-
tory and Field Measurements. J. Air Pollut. Control Assoc.
24(12):1172-1176, 1974.
5. Rao, A.K. An Experimental Study of Inertial Impactors.
Ph.D. Dissertation, University of Minnesota, Minneapolis,
1975.
215
-------
6. Dzubay, T.G., L.E. Hines, and R.K. Stevens. Particle Bounce
Errors in Cascade Impactors. Atmos. Environ. 10(3):229-
234, 1976.
7. Natusch, D.F.S., and J.R. Wallace. Determination of Air-
borne Particle Size Distributions: Calculation of Cross-
Sensitivity and Discreteness Effects in Cascade Impaction.
Atmos. Environ. 10(4):315-324, 1976.
8. Gushing, K.M., J.D. McCain, and W.B. Smith. Experimental
Determination of Sizing Parameters and Wall Losses of Five
Commercially Available Cascade Impactors. Annual Meeting,
Air Pollution Control Association, Paper No. 76-37.4, 1976.
9. Lundgren, D.A. An Aerosol Sampler for Determination of
Particle Concentration as a Function of Size and Time.
J. Air Pollut. Control Assoc. 17(4):225-229, 1967.
216
-------
PAPER 13
FIELD TESTING OF AUTOMATIC PIEZOELECTRIC MICROBALANCES FOR
OUTDOOR AEROSOL MASS CONCENTRATION MEASUREMENTS
KAZUO TSURUBAYASHI
HAJIME KANO
NIHON KAGAKU KOGYO CO., LTD.
ABSTRACT
The main purpose of this paper is to describe the perform-
ance, 'field-testing data, and sampling technique of an automati-
cally operated piezoelectric microbalance particle mass concen-
tration monitor (automatic piezobalance) which has high sensi-
tivity for particle detection. A further application, real-time
particle size measurement using an Andersen cascade impactor,
is also described.
Results presented here show a high correlation coefficient
(greater than 0.9) between the automatic piezobalance and low
volume sampling and filter weighing (LV) measurements when the
sampling conditions are set up to avoid wind effects and to mini-
mize wall losses of particles in the sampling tube.
INTRODUCTION
The Japanese law for ambient air quality standards specifies
that:
1) Regulations must be established based upon the mass
concentration of respirable particulate matter (defined
as airborne particles smaller than approximately 10 urn),
2) The standard value of respirable particulate matter
for the environment should be less than 0.1 mg/m3, when
one-hour values are averaged over a 24-hour period,
and less than 0.2 mg/m3 for each measured one-hour
value.
3) Low volume sampling, using a 10 ym cut-off device and
a filter weighing method, should be the standard mea-
suring method.
217
-------
Light-scattering dust meters have been accepted for measuring
one-hour values only if they have been first calibrated at each
local monitoring station.
Monitoring and surveillance of ambient air quality is being
done by the National Air Monitoring Network and city or prefec-
tural monitoring stations. There are more than 1000 such sta-
tions in Japan. However, ever since the law was enacted, the
standard measuring method has had two main problems that still
remain unsolved:
1) One-hour values of mass concentration cannot be measured
by the LV method, and
2) Light-scattering measurements do not correlate well
with LV measurements.
A wide range of people have requested improvement of the
standard measuring method. In 1976, the automatic piezobalance
(Piezobalance is a trademark registered in the USA by TSI In-
corporated) was developed to solve these problems. The Japanese
Environmental Agency has field tested several automatic piezo-
balances and several beta absorption instruments as alternative
measurement methods.
PRINCIPLE AND OPERATION
The automatic piezobalance, shown schematically in Figure
1, consists of a mass-sensing/auto-cleaning system and a digital
processing system. The mass-sensing/auto-cleaning system in-
cludes a vacuum pump with one 5,/min sonic nozzle, an automatic
crystal (sensor) cleaner, an analog sequence controller, a pre-
cipitator with removable precipitation needle, a detecting crys-
tal with oscillator and mixing circuit, and an impactor. The
digital processing system includes a timer-controller, LED dis-
play, and a printer that prints data, time, mass concentration,
and mixed crystal frequency.
The aerosol sample is drawn into the 10-ym impactor at a
flow rate of 1.0 8,/min by a vacuum pump. The impactor removes
aerosol particles greater than 10 vim (aerodynamic particle diam-
eter) . Particles smaller than 10 ym are carried by the air to
an electrostatic precipitator which deposits them onto a mass-
sensing piezoelectric crystal oscillating at its natural fre-
quency. The mass of particles on the crystal sensor causes the
oscillation frequency to decrease by an amount proportional to
the mass of particles. Every 10 seconds, when the instrument
is operating in the check mode, the frequency change is detected
by a counter and displayed on the digital readout. Every 120
seconds, when the instrument is operating in the measurement
mode, the frequency change from the starting frequency is con-
verted by a digital processor to mass concentration in units
of yg/m3 and displayed on the readout as shown in Figure 2.
218
-------
to
r
i
Sampling air
_
IMPACTOR
<
>
.. Precioitator
& crystal
TF
I
it
»
rixh
Cleaning I S\
Mechanismj Fllter VX
© 0-
r!
f
Son
^r
r
i
i
i
L
c
H.V.
OSC.
j
r
Processor
I/O
nozzle
-)
1
1
1
Mass sensing and auto-cleaning
Digital processor.
Printer
Figure 1. Schematic diagram of automatic Piezobalance.
-------
CLEANING FOR TWO MINUTES
fmax
-/ .
MEASURING FOR 28 MINUTES
NOT MEASURING
to
10
o
CLEANING & PRINT OUT
Cj = a(fi-fo)tx
fi: frequency shift at time tx
fo: starting frequency
a: mass concentration coefficient
Ci_i4 : average concentration in 0-28 min
C2-14 : average concentration in 30-58 min
Chr = (C1-14 + C2-14)/2
Cnr : one hour value in 0-60 min
Figure 2. Measuring system of automatic Piezobalance.
-------
The concentration C. at each sampling time t. is expressed
by x x
Ci = t. « <
where
f. is the frequency at sampling time t.
f is the starting frequency
a is the mass concentration coefficient.
The displayed concentration then is the average value of all
120-second measurements from the start until the end of the
sampling time.
After a sampling time of 28 min, a cleaning command signal
is emitted. The last 2 min of the 30-min cycle are used for
cleaning and drying the crystal sensor.
The one-hour concentration value is the average of the
values obtained during the first and the second 30-min periods
(28 min and 58 min after t^ = 0) . The one-hour value is cal-
culated automatically as defined by Equation 2:
Chr = (C1-14 + C2-14)/2
Cl-14 is the avera9e concentration during t. =0-28 min
C2-14 is the avera9e concentration during t. =30-58 min
C^r is the one-hour value during t. =0-60 min
The time and concentration are printed every 30 minutes.
After the printout, the automatic cleaning cycle begins.
The crystal is first washed with a detergent, then rinsed and
dried. After 2 min, the crystal is dry and measurement begins
automatically.
When something goes wrong during a measurement cycle, such
as loss of precipitator current, an alarm lamp turns on and the
instrument automatically stops the measurement to avoid record-
ing erroneous values.
CALIBRATION
The crystal mass concentration coefficient, a, in Equation 1
could be calculated precisely from the basic theory of piezo-
electricity if we assume ideal particle collection and sensing
221
-------
efficiency and if the particles are assumed to be deposited uni-
formly over the entire vibrating surface of the crystal.1'2 How-
ever, the automatic piezobalance requires experimental calibration
for an accurate determination of the coefficient a for the fol-
lowing reasons:
1) Aerosol particles are not deposited in a uniform layer
over the entire electrode, making the mass sensitivity
and particulate deposition profiles important.
2) Particle collection and sensing efficiencies, though
repeatable, are not precisely known for each instrument,
each precipitator needle, or each type of airborne par-
ticle.
3) The shape and size of the crystal electrode and the
thickness of the crystal are not precisely known for
a given sensor.
Therefore, we have set up a calibration procedure to experi-
mentally determine the value of a for a standard aerosol similar
to outdoor airborne particles. Figures 3 and 4 show the systems
used to calibrate the automatic piezobalances with smoke, KC1,
and PbCl2 aerosols.2'3 The basic technique is to adjust the piezo-
balance sensitivity until its concentration measurements agree
with low volume filter concentration measurements while both
are sampling from a common aerosol source.
The aerosol used for calibration is drawn through an impactor,
usually with 10-ym cut size, and into a manifold where the aerosol
is mixed and electrostatically neutralized. Several automatic
piezobalances and two low-volume filter samplers, arranged nearly
symmetrically around the manifold, sample simultaneously from
the manifold. No significant variation has been found between
sampling ports. The automatic piezobalance flow rate (1.0 &/min)
is maintained constant by its pump. The low-volume filter flow
rate (15 &/min) is continuously monitored by the operator and
maintained constant by manual adjustment. The 47-mm filters
in the low-volume sampling systems are either silver membrane
or glass fiber filters in a tight filter holder. The filter
material does not gain or lose significant weight as the relative
humidity changes. Since the particles deposited on the filter
usually do gain or lose moisture as relative humidity changes,
the filters are weighed immediately after the sampling period
at the existing relative humidity in the sampling room. The
weighing uncertainty of the gravimetric microbalance is about
±5 pg. The sample period is 0.5 - 2.0 hr, collecting approxi-
mately 100 - 3000 yg of particles on each filter. The average
of the automatic piezobalance measurements, made once every 2
minutes throughout each filter sample, is then compared with
the corresponding low-volume filter measurements.
222
-------
TOP VIEW
THERMOMETER
BALL
VALVE FILTER FLOWMETER
REGULATOR
VALVE
AEROSOL INLET PORT
PIEZOBALANCE
EXHAUST RETURN PORT
VACUUM
AUTOMATIC
PIEZOBALANCE
AUTOMATIC PIEZOBALANCE
PRE-RUN
SUCTION PORT
VALVE FILTER FLOWMETER VALVE
'!'v
' <•__ S
CHAMBER
\u
SIDE VIEW OF MANIFOLD
MIXING FAN
30LPM
J}
PIEZOBALANCE
• EXHAUST RETURN PORT
10 Jim IMPACTOR
,' \. I If ^ y
' . iTt . V
>*
-------
AUTOMATIC PIEZOBALANCE
FILTERED DILUTION AIR
FUME GENERATOR
Figure 4. Calibration system of Piezobalance by fume aerosols.
-------
Figures 5 and 6 show the result of mass sensitivity calibra-
tions obtained using the system shown in Figures 3 and 4.
Figures 5 and 6 show that the experimental sensitivity a
for the tested aerosol is within ± 10% of the theoretical value
calculated by assuming ideal particle collection, deposition,
and sensing. The data also show good linearity up to concentra-
tions at least as high as 1500 yg/m3.
40 -
30 -
GO
LLJ
O
CQ
O
N
nc
LLJ
CQ
5
D
20 -
10 -
10 864202 468 10
MASS SENSITIVITY ERROR, % (+)
Figure 5. Experimental result of mass sensitivity test.
225
-------
FIELD TESTING
After finishing the sensitivity calibrations in the labora-
tory, a pair of automatic piezobalances with mass concentration
coefficients adjusted to the theoretical sensitivity (180 Hz/yg)
were tested at two local pollution monitoring stations in Japan.
The main purposes of the field test were:
1) To evaluate the variation in measurements between two
automatic piezobalances for one-hour values and 24-hour
averaged values.
1500
CO
.E
O)
LU
o
CO
O
N
LU
o
IT
I-
LU
O
2
O
O
LU
_i
O
1-
DC
1000
500
O AUTOMATIC
A PORTABLE
O A PbCij FUME
O A KCI FUME
O A SMOKE PARTICLE
500
1000
1500
FILTER CONCENTRATION,Atg/m3
Figure 6. Experimental result of mass sensitivity calibration.
226
-------
2) To evaluate the correlation between the mass concentra-
tion measured by LV method and the concentration mea-
sured by automatic piezobalances.
3) To evaluate possible variations in sensitivity for par-
ticles at each local pollution monitoring station.
4) To evaluate the resolution necessary to accurately
measure mass concentration, within a one-hour period.
5) To evaluate the effects of atmospheric conditions,
especially relative humidity.
Figure 7 shows the experimental setup for the automatic
piezobalances at the pollution monitoring station. Two automatic
piezobalances were placed in the measuring room beside other
pollution measuring instruments, i.e., S02 monitors, NOX monitors,
light scattering dust meters, etc. Two low-volume samplers were
set on the roof of the measuring room near the inlet of the air
sampling manifold.
L.V. SAMPLER 1 L.V. SAMPLER 2
EXHAUST AIR
SAMPLING
AIR
f f v
S02 METER
NOx METER etc.
WULCL
LIGHT SCATTERING
DUST METER
AUTOMATIC PIEZOBALANCES
POLLUTION MONITORING STATION
Figure 7. Experimental set up at the pollution monitoring station.
227
-------
Airborne particles were sampled at a flow rate of approxi-
mately 150 &/min. The inlet of the air sampling manifold was
located more than 10 m above ground level. The air velocity
at the sampling points to all the instruments was about 0.5 m/s.
Polyethylene tubing with 0.375-inch I.D. carried the sampled
air between the manifold and the automatic piezobalances. A
branch fitting near the automatic piezobalances split the air
sample to the instruments.
The experimental results are shown in Figures 8, 9, and 10.
Figure 8 is the correlation between LV measurements, which sampled
the air for two or three days, and the average automatic piezobalance
value for the corresponding sampling time. Figure 8 shows that
the correlation coefficient is greater than 0.9 at every monitor-
ing station. It also shows that the theoretical sensitivity
coefficient (180 Hz/yg) used for all instruments was valid within
± 20% at every monitoring station.
CO
D)
a.
z
LJJ
O
2
O
O
LLJ
O
CD
O
N
UJ
120
100
80
60
40
20
0
A
•
Y =
7 =
y =
7 =
y =
7 =
1.08x + 4.76
0.940 ( KAN AGAWA)
0.935x + 4.39
0.984 (SUITA)
0.954x- 1.53
0.908 (SAPPORO)
20
40
60
80
100
120
140
LOW VOLUME FILTER CONCENTRATION,
Figure 8. Comparison of L V. filter and Piezobalance concentrations.
228
-------
Figures 9 and 10 show the relationship between two automatic
piezobalances for 24-hour averaged values and one-hour measure-
ments, respectively. The indicated difference between two auto-
matic piezobalances was ± 5 ug/m3 ± 10% of measured value for
one-hour measurements and ± 5 yg/m3 ± 5% for 24-hour averaged
values. The data also shows that the automatic piezobalance
has high resolution, better than 10 yg/m3, for one-hour mass
concentration measurements of ambient atmospheric aerosol in
actual field situations.
Daley et al.1* have investigated experimentally the effect
of humidity changes on the indicated mass deposit for several
CO
§
Z
O
cc
I-
o
•z.
o
o
CM
6
z
LLJ
O
CO
O
N
UJ
100
80
60
40
20
20
40
60
80
100
PEIZOBALANCE NO. 1 CONCENTRATION, j/9/m3
Figure 9. Comparison of two automatic Piezobalances for 24 hour averaged values.
229
-------
aerosol particles. Figure 11 shows the experimental setup for
measuring the humidity effects at the field testing stations.
A diffusion aerosol dryer, which can reduce humidity from 80%
to 40% RH, was installed at the inlet of one of the automatic
piezobalances. Measurements were made at 24 to 35°C, 70 to 92%
RH. The results of the experiment are plotted in Figure 12.
As shown, in the concentration range below 60 ng/m3, there is
no significant difference between the two automatic piezobalance
indications. However, at concentrations greater than 60 ug/m3,
the indication of the automatic piezobalance without the dryer
increases compared with the automatic piezobalance with the
dryer.
01
O)
zf
O
Ill
O
O
O
CM
d
2:
uj
o
CD
O
N
160
140
120
100
80
60
40
20
20
40
60 80 100 120 140 160
PIEZOBALANCE NO. 1 CONCENTRATION,
Figure 10. Comparison of two automatic Piezobalances for one hour measurements.
230
-------
CONSIDERATION OF THE SAMPLING SYSTEM
The automatic piezobalance was designed for stationary opera-
tion at fixed pollution monitoring stations.5"7 It has a sampling
flow rate of 1.0 fc/min. For the airborne particle measurements,
we must carefully design the sampling system to avoid wind-induced
losses near the inlet of the sampling tube and to minimize wall
losses in the sampling tube.
Figure 7 shows one of the most common sampling systems used
at monitoring stations in Japan. In this case, it is necessary
to use the shortest possible length of horizontal sampling tube
between the sampling manifold and the instruments to avoid the
above mentioned particle losses.
10 L CHAMBER
THERMOMETER
HYGROMETER
PIEZOBALANCE
Figure 11. Experimental set-up to check the effect of humidity.
231
-------
Figure 13 shows the relationship between the horizontal
length of sampling tube in centimeters and the particle deposi-
tion rate in percent. The calculation was made for a flow rate
of 2 £/min with 1/2-in. I.D. tubing for various particle diameters
(in micrometers) using an equation described by Pich.8
During the field tests, especially on windy days, there
was a large difference between LV measurements and automatic
piezobalance measurements. Deposition losses in the sampling
tube can account for these observed differences.
£E
O
I-
D
O
I
CN
d
LLJ
O
00
O
LU
51
100
80
60
40
20
20
40
60
80
100
PIEZOBALANCE NO. ^ (WITH DRYER), jug/m3
Figure 12. Effect of humidity on Piezobalance.
232
-------
Figure 14 is a schematic diagram of the isokinetic total
and respirable particle sampler (Isokinetic TR Sampler). This
sampler was designed to avoid the losses noted above. The sampler
consists of a sampling nozzle; an impactor which can deposit
large particles on a filter for weighing; a filter holder with
a hole in the center to allow part of the respirable particles
to collect on the filter and part of the same sample to be di-
rected to the automatic piezobalance.
This sampler can also be used for dynamic calibration to
measure precisely the mass concentration coefficient for actual
airborne particles at any monitoring location. Figure 15 shows
a dynamic calibration setup using this sampler. As shown in
the figure, one can supply test aerosol to the TR Sampler and
automatic piezobalances either from a standard particle generator
or from outdoor air. Figure 16 shows the data for a dynamic
calibration with a particle generator. The data show a very
high correlation between LV measurements and the measurements
made with the automatic piezobalances.
100
5?
uT
K
DC
Z
O
I 50
Q.
LU
Q
UJ
O
8pm
7/im
6/im
100
200
400
HORIZONTAL LENGTH OF SAMPLING TUBE, cm
Figure 13. Relationship between tube length and particle deposition rate.
233
-------
to
CO
MESH
AEROSOL INLET
1
rv.
F
\
LJ.
.*-
±=
L
^x
1
k
/
^
MBMMM
1 LP^
FILTER
^ FILTER
- 20 LPM
>
I
/I
V.P.
100
o
^
oc
50
I i i I i . • . I
0.5 1 5 10
PARTICLE DIAMETER, Aim
30
PI EZOBALANCE
Figure 14. Schematic diagram of Isokinetic TR Sampler.
-------
OTHER APPLICATIONS
When studying hazardous effects to the human body caused
by respirable particles and controlling the source of these
aerosol pollutants, it is becoming more important to measure
the mass concentration and size distribution of airborne par-
ticles smaller than 10 ym. At present, it takes one to two weeks
to get enough samples in certain size ranges to accurately weigh
using an inertial sampler such as a cascade impactor.9
An automatic piezobalance can make near real-time measure-
ments of the mass concentration within size ranges determined
by an inertial sampler. The experimental setup for measuring
the real-time mass concentration as a function of particle size
is shown in Figure 17. An Andersen ambient air sampler, modified
by making a sampling hole in the side wall of each stage, was
used for sizing the particles. Airborne particles were sampled
by the automatic piezobalance through the side wall sampling
tube at each stage of the Andersen sampler at a flow rate of
1.0 £/min.
OUTDOORS
VACUUM
PUMP
SAMPLING TUBE
CHAMBtIR
TR
SAMPLER
AEROSOL GENERATOR
PIEZOBALANCE
AIRBORNE
PARTICLES
Figure 15. Dynamic calibration set-up using TR Sampler.
235
-------
n
O)
3.
1500
tsj
U)
cc
H
Z
LU
o
I
LU
0
1000
CQ
O
N
500
500
Ser. No.
..... 8G473,y = 1 .06x - 8
O ..... 86474,Y = 0.9x + 71
I >
= 0.997
- 0.994
1000
1500
2000
LV. FILTER CONCENTRATION,
Figure 16. Dynamic calibration result with particle generator.
-------
Before the experiment, the cut-off characteristics of each
stage of the Andersen sampler were tested using 0.5 to 16 ym
oleic acid particles generated by a vibrating orifice monodis-
perse generator and the concentration was measured using a piezo-
balance portable aerosol monitor. Figure 18 shows that experi-
mental results agree well with the designed penetration charac-
teristics.
Figure 19 shows mass concentration measurements made both
by the automatic piezobalance/Andersen sampler system and by
weighing the loaded filter from each stage of an Andersen sampler
after one week of sampling. We did not find evidence of a bi-
modal size distribution in this experiment, but we did obtain
similar results using the Andersen sample and the automatic
piezobalance system.
The ratio of coarse to fine particles is shown in Figure 20
for a three-day period. This ratio can be very easily measured in
real time by the combination of the modified Andersen sampler
and automatic piezobalance.
AMBIENT AIR
VACUUM
PUMP
\
AUTOMATIC
PIEZOBALANCE
AUTOMATIC
PIEZOBALANCE
AUTOMATIC
PIEZOBALANCE
AUTOMATIC
PIEZOBALANCE
ANDERSEN AIR
SAMPLER
AUTOMATIC
PIEZOBALANCE
Figure 17. Experimental set-up for measuring the real-time sized particle
mass concentration.
237
-------
SUMMARY
Field testing data for measuring the mass concentration
of airborne particles and a newly-developed application of the
automatic piezobalance have been described. A summary of the
results is as follows:
1) The automatic piezobalance can measure one-hour values
of mass concentration within ± 20%, as compared with
LV measurements.
2) The automatic piezobalance can operate for over one
week without maintenance.
100
50 ' ~
cc
HI
0.1
0.5 1 5
PARTICLE DIAMETER, )um
10
Figure 18. Penetration efficiency characteristics of Andersen air sampler.
238
-------
3) A newly developed isokinetic total and respirable sampler
is useful for making dynamic calibrations under field
conditions and also for minimizing wind effects and
wall losses of particles in sampling tubes.
4) The real-time measurement of mass concentration by size,
in the 0.5 to 20 vim range, can be easily made using
a modified Andersen ambient air sampler and an automatic
piezobalance.
CO
a.
z"
o
<
O
o
CO
co
20
10
P'
I
-~H
i
I
I
PIEZOBALANCE
ANDERSEN, FILTER
0.1
0.4
-n
1 3 5
PARTICLE DIAMETER,
10
11
Figure 19. Size distribution measurements of airborne particles.
239
-------
to
£*
o
TEMPERATURE (OC) HUMIDITY (%)
DC
O!
01
V)
cc
<
8
2.0
1.5
1.0
0.5
\
\d
RELATIVE HUMIDITY
TEMPERATURE /
COARSE/FINE
40 1100
30 -
20
10 -
0 J 0
75
50
25
18 24
12 18 24 6 12 18 24 6
12
TIME, hr
Figure 20. Result of the real time measurements of the coarse/fine ratio
by Andersen sampler and automatic Plezobalance.
-------
REFERENCES
1. Sem, G.J., and K. Tsurubayashi. A New Mass Sensor for Respir-
able Dust Measurement. Am. Ind. Hyg. Assoc. J. 36:791,
1975.
2. Sem, G.J., K. Tsurubayashi, and K. Homma. Performance of
the Piezoelectric Microbalance Respirable Aerosol Sensor.
Am. Ind. Hyg. Assoc. J. 38:580, 1977.
3. Homma, K. Calibration Method of Mass Sensitivity for Air-
borne Particle Mass Monitor. Japan Soc. Air Pollut., 1978.
4. Daley, P.S. The Use of Piezoelectric Crystals in the Deter-
mination of Particulate Mass Concentration in Air. Ph.D.
Thesis, University of Florida, Gainesville, 1974.
5. Tsurubayashi, K., K. Homma, et al. On the Piezobalance
Airborne Dust Meter. Japan Soc. Air Pollut., 1976.
6. Tsurubayashi, K., et al. Measurement of Airborne Particu-
late Mass Concentration by Automatic Piezobalance Dust
Monitor. Japan Soc. Air Pollut., 1977.
7. Tsurubayashi, K. Measurement of Airborne Particulate Mass
Concentration by Automatic Piezobalance. Japan Soc. Air
Pollut., 1978.
8. Pich, J. Theory of Aerosol Filtration. In: Aerosol Science,
C.N. Davies, ed. Academic Press, London, 1966.
9. Tsurubayashi, K. The Real-Time Measurement of Airborne
Particulate Mass Concentration by Sizing. Japan Soc. Air
Pollut., 1978.
241
-------
PAPER 14
A NEW REAL-TIME ISOKINETIC DUST MASS MONITORING SYSTEM
JAMES C.F. WANG
COMBUSTION RESEARCH DIVISION
SANDIA LABORATORIES, LIVERMORE
ABSTRACT
A new real-time dust mass monitor is being developed by
combining an automatic isokinetic sampling probe with a tapered-
element oscillating microbalance (TEOM). Particulates from a
stack are sampled on line through the isokinetic sampler and
collected on an astroquartz mat filter of the TEOM detector.
The filter is originally excited and oscillated at a low fre-
quency (about 200 Hz). As the particulates deposit on the fil-
ter, the mass increase of the filter is reflected in a frequency
reduction which is measured in real time and yields directly
the collected particulate mass. The TEOM detector normally has
a high mass resolution (10~9 g) and wide dynamic range (10s - 106).
It has been desensitized for high particulate loading applica-
tions. The integrated monitoring system was tested in a room-
temperature wind-tunnel flow with externally injected particulates.
Good agreement was obtained between the mass collected through
the isokinetic sampling system and the weight loss of the dust
feeder in real time.
INTRODUCTION
Particulate emission control is one of the most challenging
tasks in the development of advanced fossil-fuel combustion sys-
tems. Size distribution, dust loading density, and other critical
physical and chemical properties of particulates in the effluents
of combustors and their associated cleanup equipment comprise
the key information needed for this development.1 However, there
has been a lack of accurate and real-time particulate diagnostic
techniques for the effluents from advanced power systems. This
has prevented efficient evaluations of the performance of com-
bustors and cleanup equipment and has made it difficult to de-
velop components in the advanced power systems to meet environ-
mental regulations and gas turbine product specifications. There
is thus an urgent need for appropriate particulate diagnostic
242
-------
instruments, especially those with capabilities of nonintrusive,
in-situ, and real-time monitoring. The difficulties which chal-
lenge the instrument designers originate from the hostile en-
vironment of most advanced fossil-fuel combustion systems.
Conventional techniques used in industry and utility com-
panies to characterize the effluents from their boilers or fur-
naces require sample withdrawal by a physical probe with sub-
sequent dilution and/or cooling. Insertion of the probe into
the flow may have significant effects on the flow field in the
vicinity of the probe.2 Dilution and/or cooling of the sample
during extraction may induce substantial particle size change
in the sample due to agglomeration and condensation of gaseous
species. Furthermore, analysis of the particulates collected
is usually performed off-line and may be biased or dubious be-
cause of difficulties associated with redispersion of the par-
ticulates in the analyzer.
The physical sampling method is, however, the only means
of providing samples of particulates for detailed analyses of
their physical and chemical properties. The lack of reliable
and unbiased physical sampling probes for high-temperature, high-
pressure fossil-fuel combustion environments causes uncertainties
in evaluation and modification of the combustors and cleanup
equipment. An automatic, relatively nonintrusive sampling system
is being developed at Sandia Laboratories as the first step
toward the development of a reliable and unbiased physical sampl-
ing system. An on-line mass detector is also being developed
to couple with the isokinetic probe to obtain real-time dust
loading measurements.
ISOKINETIC SAMPLING SYSTEM
The process of physically extracting a sample from a flow
system inherently disturbs the properties being measured. One
of the major sources of error in sampling is the discrimination
error of particle size due to the probe intrusion in the sampled
flow (aerodynamic error). If the sampling velocity is greater
than the undisturbed flow velocity due to heavy suction, smaller
particles are extracted preferentially since they follow the
gas stream lines more readily. On the other hand, if the flow
must decelerate on entering the probe, the larger particles
continue moving into the probe as a result of their inertia,
thus increasing their concentration. If the sampling velocity
is matched to the free-stream velocity, i.e., isokinetic sampl-
ing, the size distribution and mass loading density in the sampled
stream are preserved.
The magnitude of sampling errors from anisokinetic sampling
has demonstrated the necessity of well-characterized sampling
conditions, especially for particles in the range of 0.5 to 100 urn
in diameter.3 In practice there can be uncertainty in obtaining
243
-------
true isokinetic sampling conditions. A common method of iso-
kinetic sampling involves measuring and matching volumetric flow
rates of the probe and the free stream. This necessitates mea-
surements of pressure and temperature at the probe inlet. For
fossil-fuel combustor exhaust flows, where velocity, temperature,
and pressure fluctuate, it becomes difficult to follow these
variations and maintain the isokinetic sampling condition. A
second method is to match the static pressures at the probe inlet
and in the free stream. "* The isokinetic sampling condition can
easily be achieved by nulling the difference between the two
static pressures via a throttling valve which controls the sampl-
ing flow rate. An automatic isokinetic sampling system has been
developed employing an electromagnetic control valve based on
the "null" principle.5 However, an automatic backflush mechanism
should then be built into the sampling system to prevent the
static pressure sensors from becoming plugged in the high dust
loading environments commonly encountered in advanced power
systems.6
The automatic sampling system developed at Sandia Labora-
tories consists of (1) a sampling probe, (2) two static pressure
sensors, (3) an isokinetic sampling controller, (4) a throttling
control valve, and (5) a flow meter. Figure 1 shows the schematic
of the sampling system. The sampling probe is made of 6.35-mm
O.D. and 4.45-mm I.D. stainless steel tubing and is shown sche-
matically in Figure 2. The sampled-flow static pressure tap
holes are located in the sampling nozzle wall at about six nozzle
diameters downstream from the nozzle tip. The free-stream static
pressure sensors are on a 3.175-iran O.D. stainless steel tube
placed above and parallel to the sampling nozzle. This sampling
nozzle and sensors configuration was demonstrated successfully
in pressurized fluidized bed exhaust environments.7 The two
static sensors are connected to each side of the Validyne dif-
ferential pressure transducer via 3.175-mm O.D. stainless tubes.
The pressure transducer is protected by shunt tubing via a
solenoid valve which is open during the backflush process. The
shunt is also used to zero the pressure transducer before each
sampling operation.
The isokinetic sampling controller is presently simulated
by using a PDPll-34 minicomputer. A hardwired electronic con-
troller will be built after the system response from each com-
ponent is identified. As shown in Figure 1, the output from
the Validyne differential pressure transducer is measured on
a Validyne Model-CD12 indicator. The output from the indicator
is then input to the minicomputer through an A/D converter.
The computer sends a control signal to the electromagnetic
throttling control valve to increase or decrease the sampling
flow depending on the input from the indicator. A zero input
from the indicator corresponds to the isokinetic sampling condi-
tion; thus no action is applied to the throttling valve.
244
-------
to
*>
Ul
1SAMPLING
\\lfir PROBE
cswooo yZZZZZZs,
•FLOW
NEEDLE
VALVE
BACKFLUSH
AIR INLET
SAMPLING
STREAM SENSOR
1
OPEN
DEL
CLOS
CLO
SHU
Figure 1. Automatic isokinetic sampling system.
-------
The static pressure sensors are periodically backflushed
with dry nitrogen to prevent blockage of the tap holes. During
the backflush, the gas flow in the sensing line may produce an
erroneous signal to the pressure transducer. Consequently,
during this period, the automatic control is interrupted by an
electronic "sample and hold" circuit. This circuit continuously
holds the signal from the controller to the throttling valve,
FREE STREAM STATIC
PRESSURE SENSOR
MEAN
FLOW
DIRECTION
SAMPLING
NOZZLE
NOZZLE INLET STATIC
PRESSURE SENSOR
SAMPLED
GAS
TO PRESSURE
TRANSDUCERS
Figure 2. Schematic of the isokinetic sampling probe.
246
-------
which is held, until the backflush is completed, at a value cor-
responding to the isokinetic flow just prior to the backflush.
The periodic backflush sequence can be overridden by the operator
at any time. This allows a continuous backflush mode of opera-
tion for very high dust loading environments. When the system
is operated in this mode, the backflush flow must be reduced
to levels which do not produce erroneous signals in the static
pressure sensors.
The automatic isokinetic sampling system is also designed
to permit manual operation. For a specific sampling nozzle inlet
cross-section, the mass flow corresponding to isokinetic condi-
tion can be computed from the measured pressure, temperature,
and velocity of the sampled flow. The sampled gas flow can then
be manually adjusted via the throttling valve to match the com-
puted isokinetic mass flow. This manual operation provides an
independent backup system for automatic operation.
REAL-TIME MASS DETECTOR
A patented real-time mass measurement—a device tapered-
element oscillating microbalance8 (TEOM)—is being adapted for
ambient air particulate loading density monitoring. In this
work, the TEOM has been modified to provide real-time dust load-
ing measurements for exhausts from fossil fuel combustors.
The active element of the TEOM consists of a tube constructed
with high mechanical quality factor and having a special taper
(Figure 3). This tube is firmly mounted at the wide end; the
other end supports an exchangeable filter which can be fabricated
from virtually any material. The tapered tube with the filter
at the free (narrow) end is set into oscillation in a clamped-
free mode between two electrostatically charged parallel-field
plates. Dust-laden air is drawn through the filter and the hol-
low tapered tube via a vacuum pump. A feedback system maintains
the oscillation, whose natural frequency will change in relation
to the mass deposited on the filter. The sensitivity and fre-
quency can be chosen at will by proper dimensioning of the oscil-
lating tapered element.
Simplified operational details are shown in Figure 3. The
tapered element is originally excited into oscillation and kept
in oscillation at its natural frequency by a feedback system
which consists of a light-emitting diode photo-transistor com-
bination. The oscillation of the tapered element is converted
into an electrical signal from the diode photo-transistor set
via the light-blocking effect of the oscillating element. The
amplified signal from the photo transistor is fed back to the
conducting surface of the tapered element. The electrostatic
charges on the tapered element are then modulated by this feed-
back signal and interact with the electrostatic field between
247
-------
SIDE VIEW
TOP VIEW
FILTER
to
>£>•
00
TO PUMP
TAPERED ELEMENT
FIELD PLATES
CONDUCTIVE
PATH TO FIBER
LED
PHOTO TRANSISTOR
DATA
PROCESSING
Figure 3. Schematic of the TEOM system.
-------
the two parallel field plates, so that a steady-state oscilla-
tion of the tapered element at its natural frequency is estab-
lished. This frequency will change as the weight of the filter
increases because of the particulate deposition. A photograph
of the detector assembly and the electronic controller is shown
in Figure 4.
The TEOM is an oscillator whose frequency can be described
with two parameters, the restoring force constant K and the ef-
fective mass m, consisting of the mass of the filter, rop, the
effective mass of the tapered elastic element, mo, and the filter
loading, Am:
m = mp + mQ + Am. (1)
The relation between these quantities is given by the expression
4lT2f2 = K
m ^ '
or
f2 = ST with Ko = ^1 • (3)
If the TEOM is initially oscillating at a frequency fa cor-
responding to mass m and exhibits a frequency fj-, after a mass
uptake, the change in mass Am can be obtained as a function of
fa, fb, and Ko; namely,
and
K
fb = in + Am '
3.
Elimination of ma by combining Equations 4 and 5 leads to
Am = KQ (l/f£ - l/fa). (6)
Note that Equation 6 is independent of the filter mass mp and
the effective mass m of the oscillating tapered element. Equa-
tion 6 is also used for calibration by measuring Am gravimetri-
cally, thus obtaining KQ.
In reality the frequency is not measured directly. Instead,
one measures the time interval T required for completing a cer-
tain number of cycles, N, of the TEOM's vibration. During this
249
-------
Figure 4. TEOM detector and its controller.
-------
gated time interval, T, a high-frequency clock (running at a
frequency f_) registers a number of counts C. This procedure
leads to measurement of the time interval with great accuracy
(5 to 6 significant figures). Since
Ti = fy (i = a or b) (7)
and
C. = T.f (i = a or b) (8)
I .L C
Equation 6 becomes
Am = iFF '
c
Thus the mass uptake Am is determined by Ko (a property of the
tapered elastic element), the chosen circuit parameters N and
fc, and the counts Ca and C]-, measured before and after the mass
uptake, respectively.
Unlike most other microbalances, the TEOM is mechanically
strong and is easily automated with data output consisting of
frequency information in the 10 to 102 Hz range. Mass deposi-
tion is monitored in real time. A dynamic operating range of
over six orders of magnitude., from micrograms to grams, is achieved,
Although the TEOM measures mass by a frequency shift, as
does a QCM (quartz crystal microbalance), there is no further
similarity. Particulate collection on a QCM can only be achieved
by impaction on the crystal surface. Particulate collection
in this version of the TEOM is through a filter whose mass is
continuously monitored by the vibration of the hollow, tapered,
elastic element.
A QCM has nonuniform mass sensitivity across its surface,
and its high oscillation frequency (10 MHz) produces a high sur-
face acceleration. This necessitates strong adhesive forces
for the particles; however, even if the adhesive forces are
strong enough to prevent particulates from breaking loose under
the surface acceleration, any mass above the immediate contact
surface will fall out of phase with the crystal oscillation,
leading to the well-known saturation effect of QCM's. After
a buildup of a few microns in thickness, any additional mass
deposition is no longer reflected in a frequency decrease, thus
restricting QCM's to a low upper limit in measurable total mass
deposition. QCM's can neither measure particulate depositions
over a thickness of a few microns nor can they measure the mass
of particles exceeding a few microns in size. Consequently,
they are totally unsuitable for measuring high concentrations
of relatively large particles such as are found in smokestacks.
251
-------
The TEOM has none of these problems. All particles of in-
terest are trapped in the filter and result in a frequency shift
which is limited only by the capacity of the filter itself.
ROOM TEMPERATURE WIND TUNNEL
The integrated isokinetic real-time dust mass monitoring
system was tested in a room temperature wind tunnel flow with
externally injected particulates. A schematic of the wind tunnel
facility is shown in Figure 5. The driver is a 5-hp low head
air blower which is used in the suction mode. The air inlet
is a 25.4-cm diameter opening through the building wall. A con-
traction cone reduces the air passage to a 10-cm diameter down-
stream from a section of honeycomb flow straightener. A T-section
with a particulate injector mounted on the side flange is con-
nected at the exit of the contraction cone. A second T-section
with the isokinetic sampling probe and sensors mounted on the
side flange is installed downstream from a 1.2-m-long straight
Pyrex-glass pipe section after the first T-section. This straight
pipe allows the injected particulates to mix with the air flow
and develop into a more uniform distribution of particulates
across the sampling cross section. The air flow after the sampl-
ing section is then turned 180° and exits to the inlet of the
air blower. The flow rate through the wind tunnel can be ad-
justed between 50 and 450 SCFM via a manual butterfly valve at
the exit of the blower. Downstream from the butterfly valve,
at about 15 exhaust pipe diameters, a pitot-static pressure type
flow meter is installed to measure the actual flow rate.
A cyclone type particulate feeder was designed, fabricated,
and tested to provide continuous and uniform injection of fly
ash into the wind tunnel for particulate diagnostic tests. The
fly-ash feed rate can be controlled via the pressure drop across
the feeder bed, and monitored on-line in real time via an elec-
tronic balance. Figure 6 shows a typical dust-feed history in
terms of the weight loss of the feeder. The nearly straight
line in Figure 6 demonstrates the uniformity of the dust feed
rate. The dust loading density in the test section of the wind
tunnel can be varied between 0.01 and 10 g/m3 using this feeder
assembly. Because of the naturally generated electrostatic
charges from the fluidized fly-ash particulates inside the feeder,
electrical grounding of the feeder was found necessary for optimum
performance and safety.
The fly ash used in the wind-tunnel tests was obtained from
cyclone catches at the exhaust cleanup system of Exxon's mini-
plant pressurized, fluidized-bed facility. Fly ash obtained
from two cyclone catches in series was roughly preclassified
at mean diameters of 1 and 5 urn, respectively. Prior to being
loaded into the particulate feeder, the fly ash was sieved at
200 mesh to break up agglomerated particles and remove large
particles. A typical size distribution of the fly ash from a
Coulter analyzer is shown in Figure 7.
252
-------
U1
U>
ISOKINETIC
SAMPLING PROBE
PARTICULATE
INJECTOR
MECHANICAL BALANCE
Figure 5. Schematic of the room temperature wind tunnel.
-------
PRELIMINARY EXPERIMENTAL RESULTS
The automatic isokinetic sampling system was tested in the
room temperature wind tunnel described above. The sampling flow
was established via a vacuum pump and measured by a Hasting mass
flow meter. A positive filter using an astroquartz mat was used
to collect the particulates from the sampling nozzle. The free
stream velocity was measured by a pitot-static flow meter at
the exhaust of the wind-tunnel blower. Using the signal from
the differential pressure transducer, the PDPll-34 minicomputer
automatically adjusted the throttling control valve on the sampl-
ing line. Isokinetic sampling condition was achieved automati-
cally (when the differential static pressure was nulled) less
than a second after the free stream flow rate was changed. The
response of the sampling system was designed to be about 10 Hz,
which is adequate for most applications in the combustor exhausts
of advanced power systems.
1.5
CO
CO
O
X
<3
LLJ
DC
O
O
1.0
0.5
0
100
200
300
400
500
TIME, sec
Figure 6. A typical history of dust feed rate.
254
-------
x
a
100
90
80
70
60
50
40
30
20-
10-
0
I I
0 0.15 0.2
S-
CUMULATIVE DISTRIBUTION
I I I I
0.3 0.40.5
PARTICLE SIZE
DISTRIBUTION
1.0 2
PARTICLE DIAMETER,
1
15 20 30 40 50
100
Figure 7. Size distribution of fly ash.
-------
Figure 8 shows the experimental results of a comparison
of the measured sampling velocity and the free stream velocity
at the zero differential pressure condition maintained by the
computer. The close agreement between the two velocities demon-
strates achievement of true isokinetic sampling conditions with
the present sampling configuration with an estimated overall
accuracy of better than 5%.
The amount of particulate matter collected at the astro-
quartz filter element was found to be generally less than ex-
pected for isokinetic sampling at uniform dust loading across
the test section. Several factors contributed to this discre-
pancy. A portion of the particulates was collected on the inside
wall of the sampling probe. Clear evidence was also found via
light scattering techniques that the radial dust loading density
distribution at the inlet of the sampling probe was not uniform
and, in fact, varied as a function of the wind-tunnel flow rate.
Figure 9 shows a typical comparison of the weight of collected
samples relative to what is expected from a uniform dust distri-
bution in a free stream at isokinetic sampling conditions. At
each flow rate, the weight ratio was nearly constant and repro-
ducible. This ratio was observed to depend upon the electro-
statics of the injected fly-ash particles and the potential of
the sampling probe. Best results appear to occur when the sampl-
ing probe is grounded. However, the influence of electrostatic
charging on dust collection needs further detailed study to
optimize the isokinetic sampling processes. On the other hand,
the size distribution of collected fly-ash particles appeared
to be similar to that shown in Figure 7.
A preliminary test of the TEOM detector coupled to the iso-
kinetic sampling system was performed in the room temperature
wind tunnel. The history of the collected particulates was re-
corded in real time. At the end of each sampling test, the filter
element of the TEOM detector was weighed and compared to the
accumulated frequency shift displayed on the recorder. Figure 10
shows the good agreement obtained between the real-time measure-
ments and the accumulated weights in typical calibration tests.
A calibration constant of 4.62 x 10~3 g/Hz was found from a least-
squares fit of the calibration data.
In Figures 11(a) and (b) are shown, respectively, photo-
graphs of the TEOM filter prior to being mounted on the TEOM's
tapered element and after it had collected about 16 mg fly ash.
Most of the fly ash collected was near the top of the astroquartz
mat as shown in Figure 11(c). Figure 11(d) shows the back side
of the same mat shown in Figure 11(c) and demonstrates that there
was no trace of fly ash penetration through the astroquartz mat.
The properties of light weight, ease of handling, and 100% col-
lection efficiency for particles 1 um and larger make the astro-
quartz mat an ideal filter material for the TEOM application.
Furthermore, the astroquartz mat was tested at elevated tempera-
ture environments and proven to be useful up to 800°C.
256
-------
Experiments on real-time dust-loading monitoring were per-
formed using the combined system of the isokinetic sampling probe
and the TEOM detector. The weight loss of the dust feeder was
also monitored in real time with the combination of a mechanical
balance and electronic scales (see Figure 1). A typical test
result is shown in Figure 12. Output from the TEOM detector
was scaled up to the total flow in the test section on the basis
of the ratio of the sampled flow to the total flow and the empiri-
cal dust distribution at the inlet of the sampling probe. Thus,
direct comparison of the dust collected to the weight loss of
the dust feeder as a function of time could be obtained. Close
agreement is shown in Figure 12 between the dust collected on
TEOM and that injected into the wind tunnel. As indicated, the
dust feed rate was changed twice. The TEOM output essentially
matches the dust feed rate and follows, with acceptable response,
changes in the feed rate. The real-time on-line dust loading
monitoring capability is demonstrated.
V,
1.1
1.0
0.9
0.8
$ $
0
10 15 20
FLOW VELOCITY, Vo, m/sec
25
30
Figure 8. Test results of the isokinetic sampling system.
257
-------
1.0r-
to
Cn
oo
0.8
0.6
W,
0.4
0.2
0
0
20
40 60 80 100 120 140
Q, I/sec
Figure 9. Dust collection test results.
-------
6SZ
CALIBRATION CONSTANT -, g/Hz x
ro
CO
s
c8
O
o
0}
o
-*,
m
o
m
O
m
O
X
&
c?
o
tvj
O
NJ
X
o
CO
CO
X
N
CO
"
-------
Sandra laboratories
(a)
(b)
Sandia laboratories
(c)
(d)
Figure 11. TEOM filters.
-------
T93
ACCUMULATED DUST FEEDER WEIGHT LOSS, g
N)
CO
a1
«•»
I
3
.a
s
Ql
1'
Q>
I
1
5'
O
I
I'
•3
m
CD
o
-------
SUMMARY
The combination of an automatic isokinetic sampling system
and the TEOM real-time mass detector has been demonstrated to
be a promising real-time particulate mass monitoring system for
high dust loading environments. The bench test model described
in this paper was designed for near room temperature and one
atmospheric pressure condition. Additional development is under-
way to extend this new monitoring system to high temperature
and high pressure applications. Automatic self-cleaning features
and/or interchangeability for the TEOM filter during sampling
processes are planned to be developed for continuous particulate
mass loading monitoring applications. A particularly useful
development which we are exploring is the incorporation of a
TEOM detector at each dust collector of a cascade cyclone train.
Such a system will permit real-time analyses of mass loading
distribution according to particle size.
ACKNOWLEDGEMENTS
The author would like to thank J. Teodoro and T. Schoeppe
for their help in building the test facility and performing the
experiments. This work is sponsored by the Heat Engines Branch,
Office of Fossil Energy, U.S. Department of Energy and is a part
of the Sandia Laboratories' Diagnostic Assessment for Advanced
Power Systems Program.
REFERENCES
1. Coleman, H.W., D.R. Hardesty, R.J. Cattolica, J.H. Pohl,
L.A. Rahn, R.A. Hill, and D.L. Hartley. Diagnostics Assess-
ment for Advanced Power Systems. Sandia Laboratories Report
SAND79-8216, 1977.
2. Vitols, V. Theoretical Limits of Errors Due to Anisokinetic
Sampling of Particulate Matter. J. Air Pollut. Control
16:79, 1966.
3. Performance and Measurements at Dust Collectors. Verein
Deutscher Ingenieure, VDI-2066 Standards, 1966.
4. Branch, M.C. Sampling From High Temperature Particle Laden
Flows. Sandia Laboratories Report SAND78-8253, 1978.
5. Ringwall, C.G. Compact Sampling System for Collection of
Particulates from Stationary Sources. EPA-650/2-74-029,
U.S. Environmental Protection Agency, Research Triangle
Park, NC. 1974.
262
-------
6. Wang, J.C.F., C.G. Ringwall, and C.M. Thoennes. A High-
Temperature, High-Pressure Isokinetic/Isothermal Sampling
System for Pressurized Fluidized Bed Application. In: Pro-
ceedings 5th International Conference on Fluidized Bed Com-
bustion, Washington, DC, Vol III, p. 326, 1978.
7. Wand, J.C.F., R.R. Boericke, and R.A. Fuller. A High-Tempera-
ture, High-Pressure, Isokinetic/Isothermal Sampling System
for Fossil Fuel Combustion Applications. In: Proceedings,
Symposium on Transfer and Utilization of Particulate Control
Technology. EPA-600/7-79-044d, U.S. Environmental Protec-
tion Agency, Research Triangle Park, NC, 1979. p. 310
8. Patashnick, H. U.S. Patent 3,926,271, December, 1975.
263
-------
PAPER 15
A NEW REAL-TIME AEROSOL MASS MONITORING INSTRUMENT: THE TEOM
HARVEY PATASHNICK
GEORG RUPPRECHT
RUPPRECHT AND PATASHNICK COMPANY
ABSTRACT
A real-time monitoring device for airborne particulate matter
has been developed on the basis of a novel mass measuring device.
This new patented instrument is called a TEOM, Tapered Element
Oscillating Microbalance (Am/m < 10~6, dynamic range > 106).
Two aerosol monitoring techniques are utilized with this instru-
ment, an impaction device (mass resolution 3 x 10~10 g) and a
filter device (mass resolution 5 x 10~8 g). The TEOM filter
unit uses exchangeable filter cartridges. Time resolution is
1 minute for the impactor TEOM and 30 minutes for the filter
TEOM. (Time resolution: time required to measure an air pollu-
tion level of 10 yg/m3 with an accuracy of 10%.) The impactor
TEOM is used to measure the particulate content of air with high
time resolution, while the filter TEOM provides an absolute mea-
surement. The filter TEOM effectively represents a real-time
absolute standard for the measurement of particulate mass con-
centration in the air.
DESCRIPTION OF THE TAPERED ELEMENT OSCILLATING MICROBALANCE,
TEOM
It cannot be over-emphasized that the TEOM is significantly
different from gravimetric and quartz crystal microbalances
(QCM's). The active element of a TEOM consists of a tube con-
structed of a material with high mechanical quality factor and
having a special taper. This tube is firmly mounted at the wide
end while the other end supports a substrate (or filter cartridge)
which can be composed of virtually any material. The tapered
tube with the substrate at the free (narrow) end is set into
oscillation in a clamped-free mode. A feedback system maintains
the oscillation whose natural frequency will change in relation
to the mass deposited on the substrate (or filter). The sensi-
tivity and frequency can be chosen at will by proper dimensioning
of the oscillating tapered element.
264
-------
The operation of a TEOM is shown in simplified manner in
Figure 1. The tapered element is kept in oscillation by a feed-
back system. The oscillation of the element is converted into
an electrical signal by a light-emitting diode-phototransistor
combination, the output of the phototransistor being modulated
by the light-blocking effect of the vibrating element.
Unlike most other microbalances, the TEOM is mechanically
strong and is easily automated with data output consisting of
frequency information in the 102 to 103 Hz region. Mass deposi-
tion is monitored on a real-time basis.
A photograph of a typical instrument is shown in Figure 2.
The unit illustrated here has a dynamic operating range over
six orders of magnitude from 10~8 g to tens of milligrams. This
is a nominal sensitivity range; greater or less sensitivity can
be achieved by adjusting the dimensions of the oscillating ele-
ment and the substrate.
AEROSOL MONITORING OPTIONS WITH THE TEOM
The TEOM can be used to measure aerosol mass concentrations
with two different particulate collection techniques, impact ion
and filtration. These two methods are illustrated in Figure 3.
Figure 3a shows the impaction method where particulates in the
air stream are impacted against a substrate (greased or ungreased)
on the TEOM. Typical mass sensitivity for an impactor TEOM is
in the order of 3 x 10"10 g. Classical impaction techniques,
however, have well known difficulties and as a result, a filtra-
tion TEOM unit has also been developed. The filter unit was
developed to represent an absolute standard for the real-time
measurements of aerosol mass concentrations. In this configura-
tion, shown in Figure 3b, a filter cartridge is placed at the
free (narrow) end of the hollow tapered element. Particulate-
laden air is drawn through the filter, and the resulting filtered
air is pumped down the hollow tube. Photographs of a TEOM filter
unit are shown in Figure 4.
THEORY OF OPERATION
The TEOM is an oscillator whose frequency can be described
with two parameters, the restoring force constant, K, and the
effective mass, m, consisting of the mass of the filter (or sub-
strate) , mF, the effective mass of the tapered elastic element,
m0, and the filter (or substrate) loading, Am.
m = HL, + m + Am (1)
r O
The relation between these quantities is given by the expression:
4TT2f2 = (2)
265
-------
SIDE VIEW
TOP VIEW
SUBSTRATE
FIELD PLATES
TAPERED
ELEMENT
,_ CONDUCTIVE
PATH TO FIBER
LED
PHOTO TRANSISTOR
77/7'/////////
L_
i
AMP
u
L
COUNTER
1
DATA
PROCESSING
Figure 1. Tapered element oscillating microbalance (TEOM) configuration (typical).
TEOM OPERATION
1. Electric field is set up between field plates.
2. Image of tapered element is projected on phototransistor.
3. Oscillation of element initiated electrically or mechanically
produces an AC voltage output from phototransistor.
4. AC voltage is amplified and applied to conductive path on
element which maintains the oscillation due to interaction
with field set up in Step 1.
5. Frequency of oscillation, and hence mass on substrate, is
determined by frequency counter.
266
-------
Figure 2. Typical TEOM; substrate disc is at top center.
267
-------
or
f2 = sr with Ko
(3)
CALIBRATION PROCESS
If the mass Am is determined gravimetrically and added to
the filter, Ko can be determined from the frequencies fj and
f2 where fj is the frequency without Am and f2 is the frequency
with the filter loading mass Am. It is
K
:2 _
mF + mo
(4)
K
2 _
~
nu, + m + Am
r O
(5)
AIR FLOW
NOZZLE
SUBSTRATE
AIR FLOW
FILTER
TEOM IMPACTION UNIT
TEOM FILTER UNIT
3a
3b
Figure 3. Aerosol monitoring options with the TEOM.
268
-------
AIR INLET
ELECTRICAL
CONNECTOR
life,*, •
FILTER
FEEDBACK
ELECTRONICS
Figure 4. TEOM filter unit.
-------
From these two equations K0 for a particular device can be cal-
culated:
Am
(6)
MASS MEASUREMENTS
If the TEOM is oscillating to start with at the frequency
of fa and exhibits the frequency ft, after a mass uptake, Am can
be obtained as a function of fa, fj-, and Ko. It is
K
f2 = -2
a m
K
fz = 0_
b m + Am
Elimination of m leads to
Am = KQ (1/f* - 1/f2)
(7)
(8)
(9)
As is evident in Equation 9 the relation between frequency and
mass uptake is not linear. Note that Equation 9 is independent
of the filter mass, mp, and the effective mass, mo, of the oscil
lating tapered element.
For relatively small frequency changes
Af_
f
«1
(10)
Equation 9 can be linearized, but this will only be an approxi-
mation.
In reality the frequency is not measured directly. Instead,
the time period, T, required for completing a certain number,
N, of cycles of the TEOM's vibration is measured. During this
gated time period, T, a high frequency clock (running at a fre-
quency fc) registers a number of counts, C. This procedure
leads to measuring the time period with great accuracy (5 to
6 significant figures). Since
N
(i = a or b)
(11)
and
Ci = Vc
(i = a or b)
(12)
270
-------
Equation 9 becomes
K
Am = -7^7
-------
(QCM's have oscillation frequencies in the megahertz region,
whereas the TEOM oscillates around 100 Hz. Surface acceleration
is proportional to f2.) Also the TEOM is uniformly sensitive
to mass across the entire substrate surface. This eliminates
the need for a small sample site, rendering the total substrate
surface area available for particle capture which, combined with
a low surface acceleration, drastically reduces saturation ef-
fects. Additionally, even masses of particles hundreds of microns
in size can be measured. The TEOM, although highly sensitive,
has a sufficiently large dynamic range, due to its low suscepti-
bility to saturation, to allow sampling for long periods of time.
Furthermore, virtually any type of substrate can be used (filters,
low Z materials, reactive surfaces, etc.).
AEROSOL MASS MEASUREMENTS WITH THE TEOM
Typical results of measurements with the filter TEOM unit
are shown in Figure 5 which exhibits aerosol mass concentration
as a function of time over one day. Data were taken in real
30
co
i
o
o
O
o
111
20
10
I I I 1 I I I I I I I I I I I i I I I I
0600 0900 1200 1500 1800 2100 2400
TIME, hr
Figure 5. Air pollution in Albany on 12/11/78 (residential area) with
the TEOM filter unit.
272
-------
time every 15 minutes with a flow rate of 5 H/min through a mem-
brane filter cartridge with a 1 ym pore size. To eliminate the
effects of humidity, a hydrophobic filter (Teflon) was used and
the air stream and the instrument were maintained at 50°C. This
particular unit had a mass resolution of 6.5 x 10
- 8
The varia-
tion of the aerosol level during the course of the day is, to
a great extent, related to the traffic pattern.
Aerosol measurement results with a TEOM impaction unit are
shown in Figure 6. A single stage impactor with a 50% cutpoint
at 0.5 ym was used. The impaction substrate was made of titanium
(ungreased) with a resulting collection efficiency of 10% cali-
brated against the filter unit as a standard utilizing the ambient
aerosol. This efficiency will vary somewhat if there are signi-
ficant changes in the aerosol size distribution compared to the
ambient distribution it was calibrated against, but the purpose
of using an ungreased substrate surface was to enable the instru-
ment to operate unattended for weeks without maintenance (greased
surfaces would eventually saturate). The flow rate was 3 H/min
0900
2100
Figure 6. Air pollution on 12/14/78 versus wind condition measured
with the TEOM impaction unit.
273
-------
and data were taken during 3-minute sampling periods. The mass
resolution was 3 x 10 10 g. As with the filter unit, this instru-
ment was also run at 50°C to eliminate humidity effects. The
results shown in Figure 6 depict an initially low pollution level
(due to snowfall and high winds), which increased significantly
as the wind direction changed and the wind speed decreased.
Results from both units as they were run side by side are
shown in Figure 7. It can be seen that, with the exception of
one peak near 23:10, the measurements showing aerosol variations
correlate well. The peaks and valleys evident with the impactor
unit are somewhat sharper due to the higher time resolution capa-
bility of that instrument.
—i—i—i i—i—I—i 1—i i—i 1—r~—i
O TEOM FILTER UNIT (10 MINUTE SAMPLING)
D TEOM IMPACTION UNIT (3 MINUTE SAMPLING)
2200
2300
2400
0100
TIME, hr
Figure 7. Simultaneous measurements with TEOM filter and impaction
unit (12/12/78).
274
-------
CURRENT DEVELOPMENT
The TEOM detectors are currently being mated with particle-
size separators such as the dichotomous sampler and cyclones.
A microprocessor is being employed to completely automate the
devices. A desensitized TEOM filter unit is also being developed
with an isokinetic sampler for utilization as an on-line stack
sampling system.
SUMMARY
The Tapered Element Oscillating Microbalance (TEOM) repre-
sents a new, unique instrument capable of real-time aerosol mass
concentration measurements utilizing two collection techniques,
impaction and filtration. The impaction unit is capable of higher
time resolution, but the filter unit eliminates uncertainties
inherent with impaction collection techniques. The TEOM filter
unit measures the mass of the collected particles and the mass,
only, independent of their composition, Z values, optical proper-
ties, shapes, or any other particle property. It can be con-
sidered a detector which represents a real-time absolute standard
for the measurement of particulate mass concentration in the
air.
ACKNOWLEDGEMENT
The authors wish to thank the Environmental Protection
Agency for partial support of this work under grant R805222020.
275
-------
PAPER 16
NEW AUTOMATED DIFFUSION BATTERY/CONDENSATION
NUCLEUS COUNTER SUBMICRON SIZING SYSTEM:
DESCRIPTION AND COMPARISON WITH AN ELECTRICAL AEROSOL ANALYZER
GILMORE J. SEM
JUGAL K. AGARWAL
CHARLES E. McMANUS
TSI INCORPORATED
ABSTRACT
During the past several months, an automated system of in-
struments for measuring aerosol size distribution in the 0.005 to
0.2 ym diameter range has become commercially available. The
system consists of (1) a multiple screen diffusion battery, (2)
a continuous flow, non-pulsing condensation nucleus counter (CNC),
and (3) a switching valve to allow the CNC to detect the concen-
tration at each battery port. The system measures urban atmo-
spheres continuously for several days without maintenance. It
requires about 4 min to measure one distribution. The sys-
tem is useful for accurate measurements in relatively dirty
atmospheres (>106/cm3) and in clean atmospheres (
-------
available instruments. The report also describes a computerized
data reduction technique for the diffusion battery. Finally,
the report describes results of several simultaneous comparisons
between two commercially available techniques for making real-
time submicron aerosol size distribution measurements: the
multiple screen diffusion battery and the electrical aerosol
analyzer.
DIFFUSION BATTERY SIZING SYSTEM
The instrumentation system used for this study included
a multiple screen diffusion battery, a continuous flow condensa-
tion nucleus counter (CNC) and a valving system to switch the
CNC inlet between the 11 exit ports of the diffusion battery.
Figure 1 shows the system schematically. This particle sizing
system is useful for diameters from 0.005 to 0.2 ym.
The multiple screen diffusion battery was first described
by Sinclair and Hoopes2 in 1975. The diffusion battery consists
of 55 four-cm-diameter stainless steel screens with 250 wires
CONDENSATION
NUCLEUS COUNTER
TSI MODEL 3020
CYCLING, TIMING,
AND RESET
CONTROL CIRCUITRY
OUTLET
VALVE |
MOTOR
i ROTARY VALVE:
! .12 PORTS IN
1 'I I PORT OUT
0.3 L/min FOR CNC
SAMPLE
EXCESS FLOW CONTROL
I SWITCHING VALVE
MODEL 3042
AEROSOL IN
4L/min
PUMP
EXCESS FLOW:
TO MAINTAIN 4L/min AEROSOL FLOW
THROUGH DIFFUSION BATTERY
DIFFUSION BATTERY
TSI MODEL 3040
Figure 1. Schematic diagram of the diffusion battery, switching valve, and
condensation nucleus counter aerosol sizing system.
277
-------
per cm. Normally, 4 8,/min of aerosol is drawn through the screens
in series. The screens are mounted in 10 groups so that
a condensation nucleus counter can measure the aerosol concentra-
tion upstream and downstream of each group. Sinclair and Hoopes2
calibrated the screen diffusion battery. The screen diffusion
battery must not be confused with the collimated holes structures
or tubular diffusion battery also described and calibrated by
Sinclair et al.3
The continuous flow CNC was described and calibrated by
Agarwal and Sem.1* The instrument first draws 300 cm3/min of
aerosol through a saturation chamber where the air becomes nearly
saturated with butanol at 35°C. The nearly saturated aerosol
then passes through a 10°C condensation tube which condenses
butanol onto the particles, growing them into supermicron drop-
lets. The droplets then pass through the sensing zone of a for-
ward scattering optical particle detector. For concentrations
below 1000/cm3, the CNC measures particle concentration by count-
ing individual particles as they pass through the sensing zone.
The measurement in this range is a primary measurement, highly
accurate, with calibration required for only the constant sample
flow rate. The lowest detectable concentration is limited only
by the time required to collect a statistically significant
sample. Concentrations as low as 0.01/cm3 are automatically
displayed on the instrument's digital readout. For concentra-
tions greater than 1000/cm3, the CNC measures the light scattered
in the forward direction by all particles simultaneously in the
sensing zone. The photodetector output voltage is thus propor-
tional to particle number concentration. Although accurate
calibration has been performed only as high as 6 x 105/cm3, this
calibration has been extrapolated above 106/cm3.
The yalving system shown in Figure 1 between the diffusion
battery and the CNC has not been previously described. The
purpose of the valving system is to allow a single CNC to sample,
in sequence, each of the sample ports of a diffusion battery.
The primary component of the valving system is a rotary valve
with 12 inlet ports and one outlet port. The outlet port con-
nects directly to the inlet of the CNC. The first inlet port
samples the aerosol entering the first stage of the diffusion
battery. The next 10 sample downstream of each of the 10 dif-
fusion battery groups (or stages). The twelfth port samples
air which has passed through a high-efficiency filter, serving
as a check on the zero level of the system. Since the CNC re-
quires only 0.3 5,/min of sample aerosol flow, an auxiliary pump
and flow meter are built into the valving system to maintain
4 5,/min through the diffusion battery.
The valving system can operate in any of three modes: manual,
one-cycle automatic, or continuous automatic. In the manual
mode, the system remains on any diffusion battery port until
the operator or an externally-supplied signal commands it to
278
-------
proceed to the next port. In the one-cycle automatic mode, the
system automatically steps through all 12 ports a single time,
then waits for a command to proceed through another single cycle.
In the continuous automatic mode, the system steps through all
12 ports as in the one-cycle mode, but then automatically con-
tinues cycling indefinitely. The length of time on each port
can be adjusted from 5 to 50 s with 25 s being a typical time
required with the sizing system described here. The valving
system may be set to skip all ports above any chosen port, a
time saving feature when measuring very small particles. The
system generates a logic level signal, one second before stepping
to the next port, which can command an external recorder to ac-
cept data from the CNC. An indicator and a recordable signal
indicate which port is being sampled at any given time.
The design of the rotary valve is important for determining
the losses of particles passing through the system. Passages
are about 4 mm diameter and 100 mm long through the rotary valve
itself with no other significant obstruction. In addition, about
400 mm of 5 mm inside diameter vinyl tubing is necessary to con-
nect the diffusion battery to the CNC. Table 1 shows the experi-
mentally measured losses in the entire valving system. The mea-
sured size distribution can be corrected mathematically for these
losses.
TABLE 1. EXPERIMENTALLY MEASURED LOSSES OF
PARTICLES FLOWING THROUGH THE SWITCHING VALVE
SYSTEM
Particle diameter, Fractional loss,
Dp, urn %
0.008
0.010
0.015
0.02
0.03
0.05
0.07
0.10
30.9
25.4
14.2
10.5
7.6
3.5
2.9
2.0
279
-------
DIFFUSION BATTERY DATA REDUCTION
The data reduction method used for the diffusion battery
was a slightly modified version of the non-linear iterative in-
version suggested by Twomey5 and adapted for use on diffusion
battery data by Knutson and Sinclair.1 The program, listed in
Appendix 1, was modified for this work by Kapadia.6 Since Knutson
and Sinclair1 describe the method in detail, this report will
only mention several characteristics.
The particle size distributions in our work had a greater
range of variability than those measured by Knutson and Sinclair1
so Kapadia6 added a test for the quality of fit of the experi-
mental data with the computed solution. After every five itera-
tions, the program calculates a chi squared value using the simu-
lated raw data which would exactly result in the computed solu-
tion and the actual experimental raw data. When the chi squared
value is less than 0.001, the program prints the results and
stops. It will also print results and stop if 105 cycles of
iteration have been completed before chi squared drops below
0.001. In the latter case, the printed value of chi squared in-
dicates the quality of fit of the computed solution to the ex-
perimental data.
The program uses the diffusion battery calibration of Sin-
clair et al.2'3 in a matrix developed by Kapadia.6 It also cor-
rects the CNC data for the detection efficiency using the CNC
efficiency data of Agarwal and Sem.1* The program is set up for
use of an input data file so large batches of input data can
be processed rapidly.
Table 2 is a typical printout of the program. The first
upper column is diffusion battery port number and the second
upper column is corresponding CNC experimental concentration.
The third upper column is the ideal computed concentration which
would exactly result in the computed distribution. The program
calculates chi squared, listed at the bottom of the printout,
from upper columns 2 and 3. The first lower column lists the
geometric midpoint diameter, in micrometers, for each size in-
terval. The second lower column lists corresponding number con-
centrations for each size interval in the form dN/d(log Dp) where
N is particle number concentration (per cm3) and Dp is particle
diameter (micrometers). The third and fourth lower columns are
surface and volume concentrations, respectively, calculated from
number concentration assuming perfectly spherical particles and
using a 45-interval size array rather than the normal 10-interval
array. The bottom line of the lower columns, labeled "totals",
is the total computed number, surface, and volume concentrations
in units of particles/cm3, (micrometers)2/cm3, and (micrometers)3/cm3,
respectively. Notice that each of these values is the sum of
the appropriate column divided by four because each decade of
size contains four equal logarithmic intervals.
280
-------
ELECTRICAL AEROSOL ANALYZER AND ITS DATA REDUCTION
The EAA has been used extensively during the past six years
for measurements of a variety of submicron aerosol size distri-
butions. In the remainder of this report, the diffusion battery
particle sizing system is often compared with an electrical aero-
sol analyzer (EAA). The EAA, its performance, and its data reduc-
tion have recently been described by Sem,7 Pui and Liu,8 and
Liu et al.9
TABLE 2. TYPICAL COMPUTER PRINTOUT FOR A SINGLE PARTICLE
SIZING RUN OF THE DIFFUSION BATTERY AND CNC SYSTEM
TSI MODELS 3040 DIFF BAT AND 3020 CNC - TWOMEY METHOD
CNC ALCOHOL 9-10-79 RUN*2
PORT
0
1
2
3
4
5
6
7
8
9
10
NO,
CNC DATA
270000,
250000.
207000.
154000.
106000.
78500.
55700.
37000.
23800.
15700.
10100.
CALC DATA
270953.
245133.
202539.
155325.
112654.
78617.
53501.
35829.
23748.
15635.
10252.
hID PT. DIA
.0042
.0075
.0133
.0237
,0422
.0750
.1334
.2371
.4217
.7499
TOTALS
DNDLGD
DSDLGD
DVDLGD
7. 132E-03
9.697E+01
4.078E+04
3.204E+05
4 , 134E+05
2.710E+05
5.885E+04
4.724E+03
3.438E+02
4.443E+01
2.774E+05
3.982E-07
1.712E-02
2.277E+01
5.657E+02
2.308E+03
4.785E+03
3.286E+03
8.341E+02
1.920E+02
7.845E+01
3.019E+03
2.797E-10
2. 139E-05
5.057E-02
2.235E+00
1 .621E+01
5.976E+01
7.299E+01
3.295E+01
1.348E+01
9.799E+00
5.208E+01
NO. OF ITERATIONS= 105
CHISQ= 1.610E-02
281
-------
Several computerized procedures have been developed for
reducing EAA data. The one used in this study uses the same
Twomey5 method described earlier for the diffusion battery.
Appendix 2 is a listing of the program as adapted to the EAA
by Kapadia.6 Table 3 shows a typical printout of the results
from the program. The printout format is the same as described
earlier for the diffusion battery printout. This program is
much faster than several previous EAA programs and appears to
result in lower chi squared values (better fits) than most other
methods.
TABLE 3. TYPICAL COMPUTER PRINTOUT FOR A SINGLE PARTICLE
SIZING RUN OF THE ELECTRICAL AEROSOL ANALYZER
TSI MODEL 3030 EAA - TWOMEY METHOD
EAA ALCOHOL 9-10-79 RUN*2A
DIA, EAA DATA CALC DATA
.0032
.0056
.0100
.0178
.0316
.0562
.1000
.1778
.3162
.5623
1.0000
3,310
3.310
3.310
3.290
2.790
2,310
1.090
.320
.100
.020
0.000
3.308
3.308
3,308
3.288
2.788
2.306
1.095
.317
.090
.027
.009
MID PT. DIA
.0042
.0075
.0133
.0237
.0422
.0750
.1334
.2371
.4217
.7499
TOTALS
DNDLGD
DSDLGD
DVDLGD
0.
0.
1 .879E4-04
2. 166E+05
2.983E+05
2.980E+05
9.479E+04
1.357E+04
1 .257E+03
3.636E+00
2.353E+05
0.
0.
1 .OSOEtOl
3.826E+02
1 .666E+03
5.264E+03
5.296E+03
2.397E+03
7.025E+02
6.423E+00
3.931E+03
0.
0.
2.333E-02
1.512E+00
1 . 171E+01
6.579E+01
1.177E+02
9.473E+01
4.937Ei01
8.028E-01
8.541E+01
NO. OF ITERATIONS= 20
CHISQ= 4.201E-03
282
-------
DESCRIPTION OF THE EXPERIMENT
As with any aerosol measurement, the delivery of the sample
to the diffusion battery and EAA must be done carefully. Par-
ticles larger than 0.2 ym can be lost in the sample tube by in-
ertial impaction and gravitational settling while particles
smaller than 0.03 ym can be lost or changed by diffusion to the
walls or to each other. All particles can be changed drastically
by evaporation or condensation of volatile components.
Figure 2 illustrates the major components of the aerosol
generation and sampling system for a set of experiments involv-
ing atomized aerosols. Compressed air was regulated, dried,
and filtered before entering the aerosol generator, a stable,
single-jet atomizer. The output of the atomizer was a stream
of droplets which could be mixed with clean dilution air before
the droplet aerosol entered a diffusion dryer. The dryer con-
sisted of a straight, 12-mm diameter tube made from stainless
steel screen and surrounded by silica gel. Water or alcohol
vapor diffused rapidly to the silica gel while the dried aerosol
particles passed through. The dry aerosol then entered a Kr-
85 electrostatic neutralizer which exposed the particles to high
SWITCHING
VALVE
MODEL 3042
h
CNC
MODEL 3020
4L/mim
EXCESS .
AEROSOL'
4L/mirf
EAA
FROM ROOM
1 ^
ELECTROSTATIC
NEUTRALIZER
MODEL 3012
EAA
MODEL 303O
D
LUTION'VO' VALVE
DIFFUSION
DRYER
MODEL 3062
-*
^
3.5
ATOMIZER
MODEL 3076
AIR SUPPLY
(CLEAN, DRY
REGULATED AIR)
MODEL 3074
_COMPRESSED
AIR
Figure 2. Schematic diagram of the aerosol generation and measurement
system for the atomized aerosols.
283
-------
concentrations of positive and negative ions, allowing the par-
ticles to attract neutralizing ions. The dry, neutralized aero-
sol then entered a mixing and damping manifold where it was avail-
able to the diffusion battery and EAA inlets. Excess aerosol
was dumped from the manifold to the atmosphere.
The diffusion battery and EAA sampled simultaneously from
the manifold, each at its normal 4 5,/min aerosol sample rate.
The sample lines were less than 40 cm long, as short as possible.
It was especially important to mix the aerosol well before it
entered the diffusion battery so that its first sampling port
sampled mixed aerosol rather than aerosol from the boundary layer
with lower particle concentration. An elbow in the sample line
within 10 cm of the diffusion battery inlet seems to provide
sufficient mixing. The EAA drew its sheath air from the room.
Data was recorded continuously on a two-pen strip chart recorder.
The diffusion battery system required about 4 minutes per com-
plete sizing run while the EAA required about 2 minutes.
Figure 3 illustrates the major components of the aerosol
generation and sampling system for a set of experiments involving
combustion aerosols. The major difference from the system shown
in Figure 2 is the replacement of the entire generator by a 1-m3
plastic bag. The bag was nearly filled first with either room
4L/miru ,
4L/mirf
PLASTIC BAG
EAA
SHEATH AIR
FROM ROOM
Figure 3. Schematic diagram of the aerosol generation and measurement
system for the welding smoke and propane aerosols.
284
-------
air or dried, filtered air. Then the combustion aerosol was
introduced, in the case of welding smoke aerosol by drawing it
into a large volume centrifugal blower and exhausting it into
the nearly full bag, and in the case of propane aerosol by in-
serting the burning propane torch directly into the bag for about
10 s. The intent was not to characterize the combustion aerosol
generation process, but rather to produce a representative aero-
sol in the bag which could challenge the diffusion battery and
EAA simultaneously. By qualitatively varying the smoke intro-
duction method, we could supply the bag with a variety of aerosol
concentrations and with number median diameters in the 0.005
to 0.2 ym range.
The experimental results below are bar graphs of the mea-
sured concentration within each size range of the two measurement
systems. The error bars represent the maximum range of variation
observed in the experimental measurements. In some aerosol decay
experiments, only one measurement was possible at a given point
in time so no error bars are shown. Usually, we made two EAA
size distribution measurements during a single diffusion battery
measurement.
In addition to the programs listed in Appendices 1 and 2,
we corrected both the diffusion battery and the EAA data to ac-
count for the exponential decay of the combustion aerosols while
the sizing runs of several minutes duration were underway.
EXPERIMENTAL RESULTS AND DISCUSSION
The first aerosol we will discuss is the dry residue par-
ticles from the atomization of clean, unused, reagent grade
isopropanol. Figure 4 shows the measured number of distributions,
The diffusion battery measured three distributions in 12 minutes
while the EAA measured six distributions during that time. Both
systems measured repeatably. The diffusion battery and EAA
measured similar distributions with a tendency for the EAA to
measure slightly larger sizes. Figure 5 shows the same data
converted to volume distributions by assuming spherical particles,
If the spheres have unit density, the vertical axis is equal
to mass concentration in micrograms per m3. Figure 6 shows the
same data on a log-normal plot. The measurements of both instru-
ments were nearly log-normal with both measuring a geometric
standard deviation of 1.85. The diffusion battery system mea-
sured a number median diameter of 0.041 ym and the EAA measured
0.048 ym. This aerosol spans the central portion of the size
range where each measurement method is most accurate. The re-
sults indicate good agreement.
Fig.ire 7 shows the results of the next set of comparative
measurements. The atomizer system generated aerosol from a 0.1%
(by volume) solution of DOP in isopropanol. Again, the results
from the two measurement methods are similar. The results are
285
-------
B 6
o
i
K>
O
a.
o
^
O
I-
2-
ui
o
z
o
(J
cc
UJ
m
I -
1 — 1 — 1 1 I 1
i -- r~ — i — i — i — i — i—
© DIFFUSION BATTERY
-- A- — EAA
[ MAXIMUM RANGE OF DATA
--A-
.004 .007 .01 .02 .04 .07 0.1 0.2"
PARTICLE DIAMETER, Dp, MICROMETERS
0.4 0.7 '-0
Figure 4. Number distributions measured by diffusion battery and EAA of
aerosols atomized from 100% isopropanol.
T
80h
O
60p
z
UI
o
o 40
o
Ul
5
O 20-
~i 1 1 1—i—i i i i
9 DIFFUSION BATTERY
--&-- EAA
I MAXIMUM RANGE OF DATA
~l 1 1—1—TTT
1--A-- 1
1 — i — l i i i i
I
&—
^S—|
""""aj
$
.004 .007 .01 .02 .04 .07 O.I 0.2 0.4 0.7 1.0
PARTICLE DIAMETER. Dp, MICROMETERS
Figure 5. Volume distributions measured by diffusion battery and EAA
of aerosols atomized from 100% isopropanol.
286
-------
99.9
99.5
99
98
a.
O
4
T
95 -
90 -
ct
ui
I 50
v>
a:
ui
m
2
5-
2-
I -
0.5-
0.2-
1 1 1—I I I I I I
©-DIFFUSION BATTERY
A-EAA
"I—Mill
i i i i i i i
i i i i i i
.02 .04 .07 O.I 0.2 0.4 0.7
PARTICLE DIAMETER, Dp, MICROMETERS
1.0
Figure 6. Cumulative number distributions measured by diffusion battery
and EAA of aerosols atomized from 100% isopropanol.
1
m
^6
_l
o
l-
o:
<
Q- 5
in
O
X
oV
4
0"
0
0
1-
<
(E
t- 2
UI
o
o
o
1
o:
UI
m
5
I
i i i i i i i i i i i i i i i ' i i i i i i i i
0 DIFFUSION BATTERY
--A--EAA
J MAXIMUM RANGE OF DATA
T 1
-
-
_
T
1
/k
^
-1-
1^
1
^
-
.
_ -
1
^ 1
1
-
-
-
-
- —& 1
|
I
1
1
|
J^
^~^^B^"^H
I 1
1 A
-OO4 .007 .01 .02 .04 .07 O.I 0.2
PARTICLE DIAMETER, Dp, MICROMETERS
0.4
0.7 1.0
Figure 7. Number distributions measured by diffusion battery and EAA
of aerosols atomized from 0.1% OOP in isopropanol.
287
-------
nearly log-normal with geometric standard deviations of 1.5 and
number median diameters of 0.10 ym for the diffusion battery
system and 0.13 ym for the EAA .
Next, we measured aerosol generated by the atomizer system
from a solution of 1.0 g NaCl in 1000 cm3 of distilled water.
Figure 8 shows the results. Both instruments measured the mode
in the distribution at the same particle diameter. However,
while the EAA results are very nearly log-normal with number
median diameter of 0.06 ym and geometric standard deviation of
1.8, the diffusion battery results are not log-normal. The dif-
fusion battery appears to "chop off" some of the particles above
0.2 ym.
We also measured aerosol generated by the atomizer from
10 g NaCl in 1000 cm3 of distilled water. Figure 9 shows the
results. Again, both instruments locate the mode correctly.
The EAA results are nearly log-normal with number median diameter
of 0.08 ym and geometric standard deviation of 1.8. The dif-
fusion battery results are not log-normal probably because of
the collection of particles above 0.2 ym by impaction and gravi-
tational settling in addition to diffusion. The data reduction
program does not consider these additional collection mechanisms.
«
^
u
V.
(0 6
UJ
_j
o
cc
£ 5
in
0
X
CL
CO
O
<]
•z.
< 3
z
o
1—
NCENTRA"
r\>
O
0
ce '
UJ
CD
5
D
z
o
l 1 l i — r-i— i 1 r
_
-
—
K
^
-
-
-
: rt-
1
1
1
1
i T
~ 1 ©
1
• 1 1 — ' ' ' • i-. A 1 i i
1 1—
. .
^ '
J-
T — i — r— n
T
IT
A .
1 Wi 1
T
i i 1 1 1
1 1 1 1 | 1 1 T-|
—
-
9 DIFFUSION BATTERY
-A--- EAA
!J MAXIMUM RANGE OF DATA
_,
1
-
-
-
—A--,
i i
i
1
l
I '
1
l _
1 — zir — i
r™"-^^^"^"?
u 1 ii fti i 1 1 rti i i
.004 .007 .01 .02 .04 .07 O.I 0.2 0.4 0.7 I JO
PARTICLE DIAMETER, Dp, MICROMETERS
Figure 8. Number distributions measured by diffusion battery and EAA
of aerosols atomized from 0.1% NaCl in distilled water.
288
-------
For the next experiment, we used the bag sampling system
to sample diluted smoke from an electric arc welding process.
Figures 10 and 11 show the results of the diffusion battery and
EAA measurements, respectively. The two runs, in each case, were
12 minutes apart. The EAA results for Run 1 show a definite
bimodal shape in the number distribution with a small combustion
mode near 0.03 jam and a larger mode in the accumulation range
around 0.2 ym. Twelve minutes later, the smaller mode has dis-
appeared and the accumulation mode has also decayed. However,
the range around 0.075 ym shows no change, probably the result
of the smaller particles colliding and growing into the inter-
mediate range. While the diffusion battery results show a hint
of similar tendencies, the data is not nearly so clear, probably
because of the limited resolution of the measurement.
The final experiment reported here was the measurement of
very small particles near the lower size limit of both measure-
ment systems. We first filled the bag with clean dry air (CNC
measured 200 particles/cm3, EAA measured nothing). Then we pro-
duced a very fresh aerosol by inserting a propane torch into
the bag for about 10 s. We then closed the bag and immediately
began measurements. Figure 12 shows the volume distributions
*
IO
E
u 6
o
h-
ca
^5
in
O
X
a.
0
o
Z
i
.004 .007 .01 .02 .04 .07 O.I 0.2 0.4
PARTICLE DIAMETER, Dp, MICROMETERS
0.7
1.0
Figure 9. Number distributions measured by diffusion battery and EAA
of aerosols atomized from 1% NaCI in distilled water.
289
-------
1 I
6-
cc
£L
O.
Q
o>4
o
Z
o
|2
UJ
o
z
o
*'
UJ
CD
RUN I, t-0-4 minutes
• RUN 4,t-l2-l6mmutes
i 1 1 | |
1
_l_
L-i
.004 .007 .01 .02 .04 .07 O.I 0.2 0.4
PARTICLE DIAMETER,Dp, MICROMETERS
0.7 1.0
Figure 10. Number distributions measured by diffusion battery of diluted
welding smoke in a closed container as the aerosol ages.
a- 5
ID
O
o
a:
I- 2
o
z
o
o
cc
UJ
CD
~1 1 l—l—I I I
_i—' ' • ' '
~i—i—i—i i i i
1 1 1—I—T~T
• RUN I , t= 0-4minutes
• RUN 4, t-12-16 minutes
I 1
-1-1-1-
.004 .007 .01 .02 .04 .07 O.I 0.2
PARTICLE DIAMETER, Dp .MICROMETERS
O.4
0.7 I.O
Figure 11. Number distributions measured by EAA of diluted welding
smoke in a closed container as the aerosol ages.
290
-------
measured by the EAA beginning just after the bag was filled,
four minutes later, and four more minutes later. Additional
measurements were made by the EAA at two-minute intervals, but
are not plotted on Figure 12 for clarity. We can clearly see
some interesting coagulation growth as the aerosol ages. In
Figure 13, we see the volume concentration observed in three
size channels of the EAA: the channels from 0.0056 to 0.01 um,
from 0.01 to 0.178 ym, and from 0.1 to 0.178 ym. The smallest
channel coagulates quickly and the next smallest less quickly.
The concentration in the largest channel increases as aerosol
accumulates in that size range, having grown from smaller sizes.
Incidentally, we extrapolated the EAA sensitivity curve of Pui
and Liu8 to obtain a number concentration in the 0.0032 to
0.0056 ym range. Although this may result in poor absolute ac-
curacy, comparison from run to run is valid.
Figure 14 shows diffusion battery results from the propane
aerosol. The decay of the aerosol is evident, but not as clear
as with the EAA. The four-minute requirement for a single sizing
run limits the diffusion battery in this experiment. Also, the
CNC does not effectively detect particles below 0.01 ym, limiting
the diffusion battery-CNC system at present to the range above
0.01 ym.
g3
to*
I
Q.
Q
£
<2
z"
o
z
UJ
o
z
o
Ul
3.
I
• RUN I ,!-0-2minutes
• RUN 3, t-4-6mmutes
• RUN 5, t-8-10minutes
-1
.004 .007 .01 .02 .04 .07 O.I 0.2 0.4
PARTICLE DIAMETER, Dp, MICROMETERS
0.7 1.0
Figure 12. Volume distributions measured by EAA of fresh propane
aerosol in a closed container as the aerosol ages.
291
-------
IO.O
7.O
10
E
•^4.0
a.
a.
o
£20
-------
CONCLUSIONS
A new aerosol size measuring system has been developed con-
sisting of a compact multiple screen diffusion battery, a con-
tinuous flow condensation nucleus counter, and an automatic
rotary valve to switch the inlet of the CNC to each of the dif-
fusion battery exit ports.
A simple computer program has been developed to convert
the data from the new sizing system into submicron particle size
distribution. The same approach has been used to calculate par-
ticle size distributions for the electrical aerosol analyzer.
In an experimental comparison of the new diffusion battery
system with an electrical aerosol analyzer, we find good com-
parison in the range from 0.01 to 0.2 urn for most aerosols tested,
Both systems locate the peak in the size distributions within
that range. The diffusion battery system combined with the new
computer program appears to cut off part of measured size distri-
butions above 0.2 urn as evidenced by its inability to correctly
measure log-normal distributions in the range around 0.2 ym.
Further work is needed on data reduction methods for the dif-
fusion battery to include the effects of inertial impaction and
gravitational settling for particles greater than 0.2 ym. The
diffusion battery-CNC system also needs further evaluation to
determine particle losses below 0.01 ym. The current system
cannot effectively measure particles below 0.01 ym because of
this difficulty. On the other hand, the EAA can measure par-
ticles below 0.01 ym in circumstances where few large particles
exist. The above example of the measurement of fresh propane
torch aerosol demonstrates the EAA's usefulness for measuring
coagulation rates for such aerosols below 0.01 ym.
ACKNOWLEDGEMENTS
We received considerable assistance with the experiment
and with data reduction from Richard Remiarz. Abde Kapadia
shared his computer programs with us soon after he developed
them. We are grateful for this generous assistance.
293
-------
REFERENCES
1. Knutson, E.O., and D. Sinclair. Experience in Sampling
Urban Aerosols with the Sinclair Diffusion Battery and
Nucleus Counter. In: Proceedings: Advances in Particle
Sampling and Measurement. W.B. Smith, compiler. EPA-600/7-
79-065, U.S. Environmental Protection Agency, Research
Triangle Park, NC, 1979. pp. 98-120.
2. Sinclair, D., and G.S. Hoopes. A Novel Form of Diffusion
Battery. Am. Ind. Hyg. Assoc. J. 36:39-42, 1975.
3. Sinclair, D., R.J. Countess, B.Y.H. Liu, and D.Y.H. Pui.
Automatic Analysis of Submicron Aerosols. In: Aerosol
Measurements, D.A. Lundgren et al., eds. University Presses
of Florida, Gainesville, 1979.
4. Agarwal, J.K., and G.J. Sem. Continuous Flow, Single-Particle-
Counting Condensation Nucleus Counter. Submitted to J.
Aerosol Sci., 1979.
5. Twomey, S. Comparison of Constrained Linear Inversion and
an Iterative Nonlinear Algorithm Applied to the Indirect
Estimation of Particle Size Distributions. J. Comput. Phys.
13:188-200, 1975.
6. Kapadia, A. Ph.D. Thesis, University of Minnesota, Mechani-
cal Engineering Department, Minneapolis, 1979.
7. Sem, G.J. Electrical Aerosol Analyzer: Operation, Mainten-
ance, and Application. In: Aerosol Measurements, D.A.
Lundgren et al., eds. University Presses of Florida, Gaines-
ville, 1979.
8. Pui, D.Y.H., and B.Y.H. Liu. Electrical Aerosol Analyzer:
Calibration and Performance. In: Aerosol Measurements,
D.A. Lundgren et al., eds. University Presses of Florida,
Gainesville, 1979.
9. Liu, B.Y.H., D.Y.H Pui, and A. Kapadia. Electrical Aerosol
Analyzer: History, Principle, and Data Reduction. In:
Aerosol Measurements, D.A. Lundgren et al., eds. University
Presses of Florida, Gainesville, 1979.
294
-------
APPENDIX 1
PROGRAM DIFFBATdNPUT,OUTPUT)
C DIFFBAT USES TUOMEY'S NON-LINEAR ALGORITHM FOR REDUCING DIFFUSION
C BATTERY DATA. THIS PROGRAM CAN BE USED FOR THE TSI SCREEN BATTERY
C FOR 4.0 LPM FLOW RATE, THE ITERATIVE PROCEDURE IS STOPPED EITHER
C WHEN CHI-SQUARE IS LESS THAN 0.001 OR THE NUMBER OF ITRATIONS IS
C GREATER THAN 100, WHICHEVER OCCURS FIRST AFTER THE FIRST 20
C ITERATIONS.
C
C THIS PROGRAM WAS WRITTEN BY ABDE KAPADIA , A PH.D CANDIDATE AT THE
C U OF M.
COMMON CALIB<12,45),KDAT,DAT(12)»CDAT(12),CLOSS<45>,
•f ITER,FLOW,DIA<45> r X > ACTDAT < 11 )
DIMENSION CALDATU1)
DIA(1)=0.00316228
DPMULT=1.154781985
DO 5 J=2.45
5 DIA DAT(J)=0.0
DAT(11)=ACTDAT(11>
CALL DIFFTWO
CALDAT<11)=CDAT<11>
DO 73 1=1,10
73 CALriAT(ll-I)=CALDAT(12-I)+CDAT(ll-I)
PRINT 78
78 FORMATdX,1 1,//1X,"TSI MODELS 3040 DIFF BAT AND 3020 CMC - TWO
+MEY METHOD1)
PRINT 81,TITLE1,TITLE2,TITLE3,TITLE4
81 FORMAT(/1X,4A10,//1X,'PORT NO.',5X,'CNC DATA1,
+ 8X,'CALC DATA1)
DO 85 1=1,11
J=I-1
PRINT 83,J,ACTDAT(I),CALDAT(I )
83 FORMAT(3X,I2,8XFF8.0,8X,F8.0)
85 CONTINUE
C
C CALCULATE VALUES FOR OUTPUT
C
DO 100 J=3,39,4
100 CLOSS(J)=(0.5*CLOSS(J-2)+CLOSS(J-i)+CLOSS(J) +
+ CLOSS(J+l)+0.5*CLOSS(J+2))*4.0
DO 110 J=3,43,4
CLOSS(J+1)=CLOSS(J)*3.14*(DIA(J>**2>
110 CLOSS(J-t-2)=CLOSS< J ) *0 . 523* ( DIA ( J)**3)
TNUM=0.0
TSURF=0.0
TVOL=0.0
295
-------
APPENDIX 1 (cont)
DO 120 J=3?43?4
TNUM=TNUM+CLOSS(J)
TSURF = TSURF + CLOSS(J+l )
120 TVOL=TVOL+CLOSS(J+2)
PRINT 130
130 FORMAT(//1X?'MID PT. DIA1,4X?'DNDLGH'?7X?'DSDLGD'?7X?'DVDLGD1,/)
DO 150 J=3?39?4
150 PRINT 160?DIA?CLOSS(J)?CLOSS(J+l)?CLOSS(J+2)
160 FORMAT<2X,F6.4?2X?3(4XrlPE9.3)>
TNUM=TNUh/4.0
TSURF=TSURF/4.0
TVOL = TVOL./4.0
PRINT 190?TNUM?TSURF?TVOL
190 FORMAT1PE9.3>)
PRINT 192»ITER»X
192 FORMAT(//1X»'NO. OF ITERATIONS= '»13»5X»'CHISQ= '.1PE10.3)
DO 195 J=3»39f4
195 CLOSS(J)=CLOSS.120?.151?,29?.3D?.47?.59?.6??
+ .77?.83?.90?.95?,97»29*1.O/
KMAX=20
JMAX=45
DO 10 J=1?JMAX
10 CLOSS(J)=AC1DAT(1)/JMAX
ITER=0
100 DO 20 K=1?KMAX
DO 30 I=1?NDAT
RATIO=0.0
DO 40 J=1»JMAX
40 RATIO=RATIO+CALIB(I»J)*CLOSS(J)*CNCEFF(J)
IF(RATIO) 30?30?60
60 RATIO = DAT(I)/RATIO-l ,0
DO 70 J=1?JMAX
70 CLOSS(J)=CLOSS(J)*(l.0+(RATIO*CALIB(I.J) ) )
30 CONTINUE
20 CONTINUE
ITER=ITER+KMAX
IF
-------
APPENDIX 1 (cont)
c
C THIS SUBROUTINE COMPUTES THE LOSS MATRIX FOR TSI SCREEN BATTERY
C FOR 4.0 FLOW RATES
C
C
SUBROUTINE TSI
COMMON CALIB(12>45)rKDAT.DAT(12)rCDAT(12)»
•f CLOSS(45)fITER»FLOUfDIA(45)rXrACTDATdl)
FLOW=4.0
60 AK1=-1.14423
CKl=-2.990605
70 DO 90 J=l»45
NSCRN=0
SLOPE=-10.0**(AK1*ALOG10(DIA(J»+CM)
DO 80 1=2.11
CALIBlI-l»J)=(10.0**(SLOPE*NSCRN))-(10.0**(SLOPE*(NSCRN+I-1>))
80 NSCRN=NSCRN+I-1
90 CALIBdl t J) = 10.0**(SLOPE*55. >
RETURN
END
C
C ACHISQ
C
C THIS SUBROUTINE COMPUTES THE CHI-SQUARE FOR THE OBSERVED
C AND CALCULATED DATA FOR THE DIFFUSION BATTERY.
C
Lf
SUBROUTINE ACHISQ
COMMON CALIBd2f45)fKDAT»DAT(12)»CDAT(12)»
+ CLOSSC45) , ITER.FLOU»DIA(45> rX.ACTDATdl )
X=0.0
TCDAT=0.0
DO 5 J=liKDAT
5 TCDAT=TCDAT+CDAT
DO 10 J=liKDAT
AF=(CDAT(J)/TCDAT)-(nAT(J)/ACTnAT(l))
IF(ABS(AF) ,LT. .001 .AND. CDAT(J) .LT. .0001) GO TO 10
X = X-KAF**2)/(CDAT( J)/TCDAT)
10 CONTINUE
RETURN
END
*RDY*
297
-------
APPENDIX 2
. PROGRAM EAATWOCINPUT.OUTPUT)
COMMON /EAADAT/PHK 11,45) rCURKll) ,CUR2<11> ,X,
+ ITER,DIA<45),EDN<10>,EDS(10),EDV(10)
DIMENSION COMCUR(ll) , SENS< 45) >PC1 < 45) ,PC2<45> ,DNI(1L) ,
+ EAACURdl ) , DNDIUO)
DATA DIA/.00316228,.00365174,.00421697,.00486968,.00562342,
•f .00649382,.00749895,.00865965,.01000001,.01154783,.01333522,
+ .01539928,.01778281,.02053527,.02371375,.02738422,.03162230,
+ .03651744,.04216968,.04869679,.05623417,.06493821,.07498948,
+ .08659650, .10000007,.11547828,.13335224,.15399227,.17782807,
+ .20535266,.23713755,.27384217,.31622800,.36517440,.42169682,
t .48696789,.56234175,.64938212,.74989477,.86596497,1.00000075,
+ 1.15478285,1.33352243,1.53992768,1.77828075/
DATA DNHI/0.0,9.52E6,4.17E5,1.67E5,8.70E4,4.44E4,2.41E4,
+ 1.23E4,6.67E3,3.51E3/
DATA JMAX,I MAX,PI/45,11,3.14167
DO 140 J=1,JMAX
TMP=IHA(J)-.0125
IF 120,130,130
120 SENS( J)=2.351E6*(DIA(J))**6.262
GO TO 140
130 SENS GO TO 230
DO 20 J=1,JMAX
PC2( J)=EAACURU)/JMAX
20 CONTINUE
TCUR=EAACUR<1)
IMAX1=IMAX-1
DO 30 I=1,IMAX1
CUR1(I)=EAACUR(I)-EAACUR<1 + 1 )
CUR1(I)=AMAXKCURKI) ,0.0)
30 CONTINUE
CUR1(IMAX)=EAACUR(IMAX)
ITER=0
KMAX=20
TCUR=0.0
DO 32 I=1,IMAX
TCUR=TCUR-fCURKI)
32 CONTINUE
35 DO 60 K=1,KMAX
DO 40 J=1,JMAX
PCK J)=PC2(.J)
40 CONTINUE
DO 55 I=t,IMAX
A = 0.0
DO 50 J=1,JMAX
A = A + PHH I, J)*PC2( J)
50 CONTINUE
IF(A) 55,55,54
54 A=CUR1(I)/A-1.0
DO 61 J=1,JMAX
PC2( J)=PC2< J)*<1.0 + A*PHKI, J) )
61 CONTINUE
55 CONTINUE
60 CONTINUE
DO 70 I=1,IMAX
298
-------
APPENDIX 2 (cont)
CUR2=0.0
DO 65 J=1»JMAX
CUR2
65 CONTINUE
70 CONTINUE
IF+PCK3) + .
IF = /2.0 + PCK J-1)+PCK
+ PCKJ + D+PCK J+2>/2.0>
IF (PCK J) .LT.1E-05) PCK J) =0.0
111 CONTINUE
PCK42) = ( .5*PCK40)+PCK41 )+PCl (42) +
+ PCK43)+PCK44)+PCK45) >
IF GO TO 220
IF(CURKI) .LT. ,001*TCUR> GO TO 220
J=I*4-2
DNI(I)=PC1=0.0
210 CONTINUE
C
r
PRINT 704
704 FORMATdX,' -,//lXf'TSI MODEL 3030 EAA - TUOMEY METHOD')
PRINT 705»TITLElfTITLE2rTITLESfTITLE4
/05 FORMAT(/lXf4A10f//2XTlniA.1>8Xf-EAA DATA'.
+ SXr'CALC DATA'/)
COhCUR(ll)=CUR2(ll)
DO 221 I=1*IMAX
CO«CUR(11-I)=COMCUR(12-I)+CUR2(11-I)
221 CONTINUE
DO 222 I=lrIMAX
PRINT 706.DIA<4*1-3).EAACUR(I).COMCUP(I)
706 FOftMAT(IX»F6.4.2<8X.F6.3>)
222 CONTINUE
DO 225 I=1.IMAX1
EDN(I)=CUR1
-------
APPENDIX 2 (cont)
PRINT 706.DIA(4*I-3> .EAACUR(I)
227 CONTINUE
CALL WRTOUT(O)
GO TO 145
300 STOP
END
C
C
SUBROUTINE ACHISQ
COMMON /EAADAT/PHKll .45) .CUR1 < 1 1 > «CUR2< 1 1 > »X.
+ ITER»DIA(45)»EDN<10) .EDS( 10) »EDV< 10 >
TCUR1=0,0
TCUR2=0.0
IMAX=11
X=0.0
DO 10 I=1.IMAX
TCUR1=TCUR1+CUR1(I)
TCUR2=TCUR2+CUR2(I)
10 CONTINUE
DO 20 I=1.IMAX
ACUR2=CUR2(I)/TCUR2
AF=(CUR1( D/TCURD-ACUR2
ABSAF=ABS .EDNUO) .EDS (10) >EDV<10)
PRINT 710
710 FORMAT 7X , • DSDLGD • r 7X , • DVDLGD* > /)
TEDN=0.0
TEPS=O.O
TEDV=0.()
DO 100 I=l»10
PRINT 711rDIA< 4*1-1) »EDN
-------
PAPER 17
AN IN-SITU LIQUID DROPLET SIZING SYSTEM
DANIEL E. MAGNUS
DAVID S. MAHLER
KLD ASSOCIATES, INC.
ABSTRACT
The development of an in-situ device for measuring the size
and concentration of liquid droplets is described. The sensing
element or probe is a hot wire approximately 5 ym in diameter
and 1 mm long. When droplets impact the hot wire, the resultant
electronic signal is a diagnostic for the droplet size. An elec-
tronic processing system was implemented to analyze automatically
the signals and determine the droplet size distribution (number
of droplets in each of 14 independent size intervals). For each
sample run, the entire distribution is determined, and subsequent
off-line processing is used to compute the droplet concentration
and/or entrained mass.
The measuring system is called the DC-2 and was developed
under the sponsorship of the EPA. The device has been used to
evaluate efficiency of demisters used to remove entrained mass
from various types of scrubbers. The device has also been suc-
cessfully demonstrated in cooling tower studies and the measure-
ment of oil droplet distributions and selected acid mists.
The technology to improve the performance of the probe and
to interface the system with a microprocessor is discussed.
INTRODUCTION
In 1974 KLD Associates, Inc. developed and successfully
demonstrated an instrument using the hot-wire principle for
measuring the size and concentration of liquid droplets in a
gas stream. The demonstration of the device included both labora-
tory and field measurements. Originally the device was designed
for the droplet size range from 20 urn to 500 ym. However, the
field studies showed the importance of measurements in the
range from 1 ym to 100 ym, where a majority of the droplets are
found. Based upon these studies, the electronics of the system
was redesigned to include droplets in the size range from 1 ym
to 600 ym.
302
-------
Under a contract from the Environmental Protection Agency,
the apparatus for the measurement of droplets was to be improved
and refined to enable in-situ measurements under adverse field
conditions. The desired improvements in this area of technology
included:
reducing the number of adjustments to be performed by
the operator and providing for automatic data acquisi-
tion,
providing 14 sub-intervals of droplet sizes over the
range from 1 ym to 600 ym,
building into the electronics the calibration curve for
droplet size and displaying the results in digital form
in engineering units, and
improving the ruggedness of the probe to make it more
suitable for field measurements.
The above design objectives were incorporated into a droplet
measuring device which is designated as the DC-2. The purpose
of this paper is to describe the results of this design effort
and to discuss both laboratory and field results achieved with
the equipment. Also presented are some recent new design
activities to interface the device with on-line computers.
PRINCIPLE OF OPERATION
The operation of the heat transfer droplet sensor is based
on the local cooling caused by a droplet attaching to a hot wire.
The concept is schematically shown in Figure 1, where the hot
wire and its longitudinal temperature distribution are shown
(a) before a water droplet (cross-hatched circle) interacts with
the wire. The electrical resistance of the wire in (a) is high
and substantially uniform along the wire. In situation (b),
the portion of the wire covered by the droplet is cooled to ap-
proximately the droplet temperature. With a constant electrical
current heating the wire, a measurable voltage drop can be sensed
between the wire supports. The voltage for condition (a) (be-
fore the droplet attachment) is reduced in direct proportion
to the cooled length of wire, i.e., the droplet diameter (con-
dition b). The electrical energy delivered to the wire evapo-
rates the water, leaving the sensor dry and ready for further
interactions with droplets.
The above description of the operating principle for the
hot wire probe is an idealization and in actual practice the
electrical signal is rather complex. A typical electrical signal
obtained during a droplet-hot wire interaction is shown in Fig-
ure 2. Reduction of the voltage implies cooling of the wire.
An initial fast decay of the signal is observed; it corresponds
303
-------
• SUPPORT
SUPPORT
•HOTWIRE
WIRE TEMPERATURE
i
HOT- •
AMBIENT
WIRE LONGITUDINAL
DIMENSION
(a) TEMPERATURE BEFORE ATTACHMENT
WIRE TEMPERATURE
HOT-
AMBIENT
WIRE LONGITUDINAL
DIMENSION
(b) TEMPERATURE AFTER ATTACHMENT
Figure 1. Principle of operation of the sensor.
304
-------
VOLTAGE
U>
O
DROPLET
WIRE
CONTACT
LONGITUDINAL
COOLING
DROPLET EVAPORATION
TIME
Figure 2. Electrical signal from sensor.
-------
to the initial contact of the droplet and the wire. During this
period the wire is cooled radially and rapidly assumes the tem-
perature of the droplet. The duration of this portion is of
the order of a few microseconds. Following the initial contact,
the signal changes more slowly as the droplet centers itself
around the wire and as longitudinal cooling of the wire takes
place. During this period of the interaction, a warming of the
droplet/wire takes place, raising the signal until the boiling
temperature is reached; the droplet shrinks due to evaporation
and then disappears. The voltage in the wire then returns to
the equilibrium level prior to the interaction of the droplet.
The foregoing concept of measuring droplet size and concen-
tration is implemented using two major components: a hot wire
probe as a transducer and electronic instrumentation for analyz-
ing the electrical signals for droplet size and storing the data,
These components are shown in Figure 3.
Figure 3. Droplet counter and probe.
306
-------
The probe is made from platinum wire 5 urn in diameter, and
the active portion of the wire is one millimeter long. With
this size wire, the DC-2 can measure diameters of droplets from
1 ym to 600 ym. As shown in Figure 3, the fine wire is mounted
across a ring at the top of the transducer, which is connected
to the instrumentation box by a single coaxial cable. This type
of probe is called an open probe since it is unprotected against
breakage during handling. Other designs have been fabricated
with sliding sleeves which cover the probe and successfully
minimize damage from handling.
The electronics box for analyzing the signals from the probe
determines the size of the interacting droplet with respect to
14 size ranges and permanently stores the count for each range.
The size ranges can be adjusted by changing a plug-in ladder
network. The DC-2 comes equipped with four ladder networks,
and thereby allows the operators to study a variety of droplet
distributions. The equipment is designed for simple operation
and a single pushbutton switch (START) automatically cycles the
DC-2. When the START switch is pushed, the device remains in
the droplet counting mode until a preselected number of drop-
lets (100, 1,000 or 10,000) interact with the probe or a pre-
scribed time limit (10, 100, 1,000 seconds) is reached. After
completing a measurement cycle, the system halts and the data
can be read from a liquid crystal display, which was chosen for
its superior performance in bright ambient light.
The electronics box has been designed to provide the field
user with a light-weight and compact instrument. The box is
14% in. long, 6 in. wide, and approximately 6 in. high and
weighs approximately 10 Ibs. The case and card bucket are made
of drawn aluminum; the circuit cards are wire-wrapped and made
of glass epoxy. The instrument is powered from 125VAC (± 15%),
60 Hz line requiring 2.5 watts of power. The electronic system
is designed using low power, CMOS logic with a maximum counting
rate of 500 droplets/sec. In the report by Medecki, et al.,
a description of the logic and circuit diagrams is presented.1
LABORATORY INVESTIGATIONS
Laboratory investigations with the DC-2 have emphasized
three types of activity:
droplet attachment mechanism
• calibration procedures
comparisons with other droplet measurement techniques
For the first two types of activity, attachment mechanisms
and calibration, carefully controlled procedures for generating
307
-------
droplets are required as well as reliable techniques for record-
ing and verifying the droplets interacting with the hot wire
probe. A variety of techniques were used to generate monodis-
persions and polydispersions of droplets. These techniques in-
cluded: saturated steam, vibrating capillary, micropump, rotat-
ing disc, Berglund Liu generator, and various nozzles. The mono-
dispersion generated with the Berglund Liu apparatus was most
useful for developing the initial calibration curve for the de-
vice. However, the available size range was restricted and there-
fore other techniques were used to extend the range of calibra-
tion. Working with a polydispersion, the interacting droplets
are observed under a microscope during attachment to the hot
wire.
A photographic camera records the image of both the wire
and the droplet, and a permanent record is obtained to study
the attachment mechanism and to provide the calibration data.
Since the interaction time is short, an electronic flash is used
to stop the motion at the instant of droplet contact; no appre-
ciable shrinkage due to evaporation takes place and accurate
calibration data is obtained. The duration of the flash is ap-
proximately one microsecond, which is compatible with movement
and shrinkage of the droplet in the diameter range of interest.
Simultaneously, the electrical signal from the probe is displayed
on an oscilloscope and photographed. These two photographic
records provide the data for calibrating the DC-2.
With such equipment, the attachment mechanisms of droplets
can be comprehensively studied. In this manner, the effects
of surface tension between the droplet and the wire have been
investigated. Further, the influence of eccentric collision2
between the droplet and the wire was studied as a function of
droplet size and velocity relative to the wire. Also the limita-
tions on flow velocity were investigated3 and an upper bound
on the droplet size was designed for a flow velocity of approxi-
mately 10 m per second for droplets no larger than 600 urn. If
higher flow velocities are encountered, the large droplets tend
to shatter upon contact with the wire.
Using the above-mentioned equipment, a complete calibration
curve has also been developed for the DC-2 and this calibration
curve has been implemented into the electronics of the system.
The calibration curve provides a direct relationship between
the droplet size and the electronic signal generated by the
probe, as droplets interact at low velocities. Under actual
field conditions, the flow influences the droplets interacting
with the wire. This type of flow phenomena is usually expressed
in terms of a capture area which is a relationship dependent
upon droplet size. Such problems have been investigated in
aerosol filtration **'5 using a two-dimensional analysis for the
flow past an infinite cylinder. It can be shown that viscous
conditions have a significant effect upon the capture area for
small droplets. These effects are most important in the size
range from 1 ym to 10 ym.
308
-------
To further verify the performance of the DC-2, a series
of laboratory tests were performed to validate the device with
other apparatus such as the Brink impactor which is used to
sample droplet and particulate distributions in industrial ap-
plications. To minimize any measurement error, a closed system
with a steady flow was used in the laboratory (Figure 4). The
droplets were generated using a disc rotating at a constant speed
in a closed box. The entrained droplets were drawn from the
box by a constant volume pump (0.03 ft3/min) located downstream
of the impactor and the flow was returned to the box. During
the test, the droplet distributions upstream and downstream of
the impactor were monitored by the DC-2s noted as A and B in
Figure 4.
The system was operated with pure water or a 10% sodium
chloride solution. Aluminum foil cups were used as targets in
the impactor. The tests with pure water identified some sources
of measurement errors in the impactor. For example, the DC-2
at location A indicated a stable droplet distribution, but the
impactor results were erratic. The length of the sampling time
has a pronounced influence on the measurements with the impactor
and the scatter of data was found to be unacceptable. These
problems were attributed to evaporation and/or condensation taking
place at various points in the impactor. Because of these prob-
lems with the impactor, the quantitative studies with pure water
were discontinued early in the research.
FLOW
MANOMETER
Figure 4. Laboratory apparatus.
309
-------
The impactor measurements with a salt solution were more
dependable, but not entirely satisfactory. Again, the impactor
results tended to have excessive scatter which was more pronounced
as the sampling time increased. By very careful measurement,
it was determined that the salt concentration of the liquid col-
lected at the impact target is less than 10% and often as low
as 5%. This result suggests that the samples were being diluted
by condensation of water vapor from the saturated flow. Another
interesting observation from the impactor studies was demonstrated
by the transient behavior at its output. The droplet concentra-
tion measured by the DC-2 at location B was initially very low.
Approximately 10 to 60 minutes (depending on the number of impac-
tor stages) were required for the droplet distribution to sta-
bilize and to establish a meaningful distribution. Hence, the
impactor must be conditioned to the sampled environment to achieve
reliable data.
All subsequent tests were carefully designed to permit ade-
quate time for the impactor to stabilize, but the impactor data
still exhibited some scatter. In an attempt to further improve
performance, the impactor was placed inside the box where the
droplets are generated. After "soaking" in the test environ-
ment, tests were performed, but only minor improvements in the
performance of the impactor were observed.
Based upon this experience with the impactor, some carefully
selected and controlled tests were performed to provide data
for comparing the DC-2 and the impactor. A typical distribu-
tion of number concentration versus droplet diameter is shown
in Figure 5. The results from the DC-2 are presented as the
dashed curve and include the effects of the flow field on the
capture size of the wire. The results from the impactor are
rather close considering the scatter associated with such mea-
surements .
Another series of tests currently in progress utilizes a
well-calibrated spray nozzle in an open system. The nozzle pro-
duces a very high concentration of droplets in the size range
from 1 urn to 50 urn. A very carefully controlled uniform flow
condition is established with this system and the flow is con-
fined to a tube approximately 1 inch in diameter. Steady state
conditions are readily established with this system and the mass
flow is determined at a specified operating pressure by direct
measurement of the consumed liquid. In this manner the entrained
mass measured with the DC-2 can be directly compared with the
fluid flow through the nozzle. These tests are currently in
progress and initial results indicate a favorable comparison.
310
-------
FIELD/LABORATORY EQUIPMENT
The basic equipment for measurements in the field and the
laboratory consists of the probe and the DC-2 electronic box.
The droplet distribution data acquired during each sample period
can be manually read via the liquid crystal display and the
results recorded on appropriate log sheets. To facilitate the
collection of many data samples, a printer/controller unit was
developed for controlling two DC-2s. The unit is useful when
simultaneous measurements are required upstream and downstream
of a device (e.g., to evaluate the efficiency of a demister).
A schematic configuration of the equipment is shown in Figure 6.
Both DC-2 units, A and B, are under the direct control of
the printer/controller. Samples can be taken in a continuous
mode with the data from units A and B printed after the comple-
tion of each data cycle. The system can also be used for periodic
sampling with units A and B. The time period can be preset to
5, 10, 15, 30, 60 or 120 minutes by the operator. In either
mode, the printed data includes all the measured parameters and
the real-time when the data cycle was completed. After complet-
ing a specified number of data cycles, the printer/controller
automatically terminates operation.
10°
— ^
E
3.
M
0
— ' 104
Q~
I
r—
o"
e io3
D
aC
E
^_
LI
J
5 1Q2
J
c
LI
D
E
^
J
Z
101,
— . IMPACTOR
DC-2 DATA WITH FLOW
Ij FIELD CORRECTION
V
" \
\
\\
\\
\ \
\\
\\
\ x
\ x^
X \
X ^
v ^N.
X * \^
N^ »^
x ^>.
x>^
t r • j
4 6
DIAMETER D, jum
10
Figure 5. Number concentrations for the DC-2 and impactor. Impactor
sample period equals 80 min.
311
-------
More recently, KLD has interfaced the DC-2 with a micro-
processor system to further facilitate the acquisition of a large
number of data sample points. This type of arrangement has been
very advantageous during the laboratory tests where hundreds
of data samples are recorded and must be subsequently processed
to analyze the results. The printer and computer for this system
are too large and bulky for application to field work. However,
KLD has recently designed a compact microprocessor system which
is capable of data acquisition and analysis in the field environ-
ment.
For application in the field, devices to support the probe
while surveying across large ducts are available. Currently
a telescoping pole approximately 2 in. in diameter is used to
survey across a large duct in typical power plant applications.
The pole is capable of extending to 14 feet. The pole itself
is supported by a support bearing attached to the sidewall of
the duct. This type of equipment reduces the data activities
in the field to a routine procedure after the appropriate access
ports have been installed.
PROBE
A
DC-2
o
o
PRINTER/CONTROLLER
DC-2
PROBE
Figure 6. Printer/controller for automatic operation with two
DC-2 units.
312
-------
FIELD MEASUREMENTS
The measurement of droplet distributions and concentrations
is important in several applications such as the evaluation of
demister performance in power plants, assessment of entrained
mass in quenching towers and environmental control in large manu-
facturing complexes where oil mist can be a problem. Some of
these studies have been documented in other reports.
The field studies at the limestone scrubber/demisters pro-
vided the most severe test environment because of the accumu-
lation of limestone on the sensor. However, the contamination
of the sensor is easily detected as a reduction of the droplet
counting rate or by periodically inspecting the sensor. Then
the contaminated sensor is cleaned with hydrochloric acid or
the entire probe is replaced to expedite the work.
Field studies of limestone scrubber/demister have been per-
formed at power plants at Shippingport, PA and Becker, MN. At
Shippingport a 12-foot square section was surveyed through six
ports approximately 40 feet downstream of the demister stage.
At one site, the tests were performed downstream of a single
demister. At the second site, two demisters were in series and
measurements were taken after the second demister. For both
sites the duct downstream of the demister is horizontal. Such
a survey required about two hours to complete following equip-
ment set-up. A total of 51 droplet distributions were recorded.
The average velocity measured with the DC-2 was 10 m/sec,
and this value was used for computing the number concentration.
Typical results are shown in Figure 7 as a function of droplet
size. The upper curve is the number distribution for the site
with a single demister, whereas the lower curve is for the double
demisters in series.
The data for the two sites is summarized below:
Number of Demisters
1 2
Volume concentration (gr/ft ) 0.21 0.093
Droplet concentration (no./cm3) 525 301
Mass mean diameter (urn) 11.9 11.0
At the Becker Power Plant the mass carry-over from the de-
mister section was determined. In this power plant the demisters
and duct work have a vertical orientation and the surveys were
performed over a cross-section of approximately 15 x 21 ft.
Five access ports were located along each of the longest sides
of the duct. The surveys were performed by extending a pole
approximately half-way into the duct and taking measurements
from the mid-point of the duct to the wall. A total of 30 spatial
locations were used over the entire cross-section. At each loca-
tion two to four measurements were taken and averaged to obtain
the distribution at each location.
313
-------
A demister with two rows of chevrons is used in this power
plant. During this evaluation period, one row of chevrons was
removed and the tests repeated to assess the effect on the mass
carry-over. The size of this coal power plant is 760 megawatts.
The scrubbers remove approximately 80% of the sulfur dioxide
and the liquid to gas ratio is 30 gallons/1000 ft3.
CONCLUSIONS
The hot-wire sensor is an effective means of measuring the
droplet size and concentration. The DC-2 has been successfully
used in the laboratory and has provided valuable insight into
the behavior and performance of impactors when used as a sampling
device for droplet distributions. Also, the DC-2 has been demon-
strated in the field under various adverse environments. The
field studies were successful and useful measurements were ob-
tained. Measurements can be taken rapidly with a minimum of
labor and set-up time.
cc
I-
LLJ
u
z
o
o
DC
LU
DQ
2
13
z
103
10°
10'
10
-2
10
,-3
10'=
10
-6
SINGLE DEMISTER SECTION
DOUBLE DEMISTER SECTION
i
i
10 100
DROP DIAMETER D, urn
1000
Figure 7. Number concentration vs. droplet diameter for a single and
double demister.
314
-------
ACKNOWLEDGEMENTS
The research and development work on hot wire technology
for the detection and measurement of droplets was sponsored by the
Environmental Protection Agency, Research Triangle Park, NC.
Mr. Bruce Harris was the Project Officer for the EPA and his
support and suggestions are gratefully acknowledged. The mea-
surements at Shippingport were coordinated by Mr. Dennis Martin
of York Research. The measurements at the Becker Power Plant
were sponsored by Combustion Engineering, Inc. and were under
the supervision of Mr. Kal Malki.
REFERENCES
1. Medecki, H., K. Wu, and D.E. Magnus. Development of Droplet
Sizing Technique for the Evaluation of Scrubbing Systems.
EPA-600/7-79-166, U.S. Environmental Protection Agency,
Research Triangle Park, NC, July 1979.
2. Medecki, H., M. Kaufman, and D.E. Magnus. Design, Develop-
ment, and Field Test of a Droplet Measuring Device. Environ-
mental Protection Technology Series. EPA-650/2-75-018,
U.S. Environmental Protection Agency, Research Triangle
Park, NC, February 1975.
3. Medecki, H. and D. Magnus. Liquid Aerosol Detection and
Measurement. Presented at 68th Annual Meeting of the Air
Pollution Control Association, 1975.
4. Davies, C.N., ed. Aerosol Science. Academic Press, New
York, j.966.
5. Fuchs, N.A. The Mechanics of Aerosols. Pergamon Press,
New York, 1964.
315
-------
PAPER 18
SOME AERODYNAMIC METHODS FOR SAMPLING INHALABLE PARTICLES
WALLACE B. SMITH
KENNETH M. GUSHING
14. CHRISTINE THOMAS
RUFUS R. WILSON, JR.
SOUTHERN RESEARCH INSTITUTE
AND
D. BRUCE HARRIS
INDUSTRIAL ENVIRONMENTAL RESEARCH LABORATORY-RTF
U.S. ENVIRONMENTAL PROTECTION AGENCY
INTRODUCTION
At present, most measurements used to characterize particulate
emissions from stationary sources are made with cascade impactors,
cyclone trains, and filters. However, these methods do not dis-
tinguish between the inhalable and non-inhalable fractions of
the suspended particles. The objective of the investigation
described in this paper was the development of equipment and
procedures for measuring the concentration of inhalable particles,
either by modifying existing sampling systems and procedures
or by developing new ones. The systems that were developed are
to bs used for establishing emission factors for stationary
sources by measuring the concentrations of inhalable particles
in stacks and in fugitive emissions. The systems were designed
to have a second cut point at 2.5 ± 0.5 urn aerodynamic diameter.
This requirement ensures that the systems used to measure the
emission factors will yield data comparable to that measured
in the ambient atmosphere using the EPA dichotomous samplers.
The devices considered as candidates for inhalable particulate
samplers were impactors, cyclones, and horizontal elutriators.
Impactors are compact and the theory describing their opera-
tion is well developed. However, impactors of conventional de-
signs are inefficient collectors for particles with diameters
larger than the cut point values of the impactor stages.1-3
Particle bounce resulting in reentrainment is evidently respon-
sible for much of the inefficiency.
316
-------
Cyclone collectors are not subject to this problem, and
they also have the advantage that the inlet is at a right angle
to the axis of the cyclone and hence the sampling nozzle can
be pointed directly into the gas stream in the stack without
the need for a button-hook or curved nozzle. On the other hand,
cyclone samplers are bulky compared to other samplers, and no
adequate theory exists for their performance.
Horizontal elutriators are also somewhat bulky and they
are sensitive to orientation. However, reentrainment is not
a problem, and the theory describing their performance is straight-
forward and accurate.1*
Four prototype systems were investigated during this study:
(1) horizontal elutriators for sampling fugitive emissions and
to be used as precollectors for in-stack cascade impactors and
the EPA Source Assessment Sampling System (SASS), (2) a high-
volume cascade impactor for fugitive emissions, (3) a precollec-
tor cyclone for use with in-stack cascade impactors, and (4)
a dual-cyclone train for in-stack sampling. All of the prototype
systems were evaluated under conditions of gas flow and tempera-
ture consistent with those present in actual sampling conditions.
EXPERIMENTAL INVESTIGATION AND RESULTS
Prototype systems employing horizontal elutriators, inertial
impactors, and cyclones were fabricated and tested in the labora-
tory.
Horizontal Elutriators
Horizontal elutriators have consistently predictable col-
lection efficiencies, as confirmed by Stein, et al.,5 Rimberg
and Thomas,6 and Corn, et al.7 Elutriators have been used both
as particle sizing instruments and as particle collection de-
vices . 8
The experiments described in this section were performed
to verify the theory of particle collection by settling and to
develop design criteria for inlet collectors to be used in con-
junction with two sampling systems developed for the Environ-
mental Protection Agency: the Fugitive Ambient Sampling Train
(FAST) and the Source Assessment Sampling System (SASS). Designs
were also developed for applications with in-stack cascade im-
pactors.
The initial concept of the horizontal elutriator was the
result of studies to measure quantitatively the deposition of
aerosol particles on surfaces adjacent to, or suspended in, mov-
ing gas streams flowing through laboratory apparatus. Natanson1*
317
-------
and Thomas6 independently solved this problem theoretically for
a circular horizontal tube of radius a. Although the deriva-
tions are somewhat complex, the results can be summarized and
simplified in terms of the collection efficiency (Eff) by the
following equation:
Eff = f [2e Vl - £2/3 - £1/3 Vl - £2/3 + arcsin e1/3] (1)
where
3V L
s
£ 8aV
V is the settling velocity,
S
L is the length of tube,
a is the radius of tube, and
V is the average velocity of the gas.
The theory describing the performance of a rectangular flat
plate elutriator is considerably less complex and is re-derived
here to give the reader a better understanding of the particle
motion in the elutriator. The settling velocity of spheres in
a gas stream is given by the equation
where
V is the settling velocity,
o
g is the acceleration due to gravity,
C is the slip correction factor,
p is the particle density,
d is the particle diameter, and
Vi is the viscosity of the gas.
In health-related studies, the aerodynamic behavior of the
particles is of interest. It is therefore convenient to relate
the settling velocity of all particles, whatever their shape
or density, to that of spheres having unit density. The aero-
dynamic diameter is defined as the diameter of a unit density
sphere having the same settling velocity as the particle of
interest. The aerodynamic diameter is given from Equation 2
by setting p = 1 g/cm3 (the units must be retained) . Also, for
318
-------
particles larger than about 2 ym, the slip correction factor
is approximately equal to unity. A graph of settling velocity
vs. diameter is shown in Figure 1.
For fully developed flow between parallel plates, the velocity
profile of a gas is parabolic with the maximum velocity equal
to 1.5 times the average and with zero velocity at the plates:
where
V is the velocity parallel to the plates, at a point y,
A
y is the displacement above the bottom plate,
h is the spacing between plates, and
V is the average velocity of the gas.
The velocity profile and a typical particle trajectory are
illustrated in Figure 2.
The efficiency with which particles are collected is deter-
mined by the dimensions of the channel, the velocity of the gas,
the settling velocity of the particles, and the height at which
the particles enter the channel.
Consider a particle which enters the channel at position
y0 as shown in Figure 3 and which has a trajectory of length
L (the length of the channel) . All particles of this diameter
or larger entering at or below position y will be captured.
All particles of this diameter entering the channel at positions
higher than yQ will penetrate the channel.
Now, from Equation 3:
at ~Vh " h2 / v
but dy is given by dy = -V dt; therefore
s
jf-'-UVr)*'
'0 vs "y
*o
(4)
319
-------
I
I 0,
Ul
LLJ
V)
0.01
0.001
0.6 1 2 4 8 10 20 40
AERODYNAMIC PARTICLE DIAMETER, urn 4181-86
Figure 1. Settling velocity in air for unit density spheres.
320
-------
v>
V
X
\ .
o
4181-106
2. Velocity profile and particle trajectory between parallel plates.
Figure 3. Zone of 100% particle collection.
4181-73
321
-------
from which the settling velocity of the particle may be deter-
mined.
If the particles are uniformly distributed within the gas,
the fractional collection efficiency of the particle illustrated
in Figure 2 is equal to the ratio of the volume of gas passing
below position yo to the total volume; thus:
Eff. =
•/
•fr\
dy
where W is the width of the channel.
Now, from Equation 4,
Eff. . V . CE^
h v 18 Mh V
The theoretical efficiency curves for a horizontal elutriator
with a circular cross-section (from Equation 1) and for a hori-
zontal elutriator with a rectangular cross-section (from Equa-
tion 5), both designed to have a cut point of 15 ym aerodynamic
diameter, are shown for comparison in Figure 4.
The efficiency is found to be independent of the vertical
velocity profile and, as Fuchs1* observed, Equation 5 can be
derived more easily assuming plug flow.
Experimental Procedures—
Experiments were conducted to verify Equation 5 and to set
design parameters for the FAST and SASS inlets.
Figure 5 is a schematic diagram of the experimental apparatus
used to investigate the performance of a horizontal elutriator
designed to have a cut point (D50) of 15 um aerodynamic diameter.
The settling chamber consists of 28 channels, each 7 mm high,
17.8 cm wide, and 38.1 or 20 cm long. A high volume blower
(Model 305, Sierra Instruments) connected in series with a vari-
able voltage transformer was used to supply the desired air flow
rate through the chamber. Monodisperse particles of methylene
blue were generated using a vibrating orifice aerosol generator
(VOAG) for particles larger than 4 ym aerodynamic diameter.
During each test, the particles were sampled and checked fre-
quently by optical microscopy to ensure constant monodispersity.
322
-------
All particles entering the horizontal elutriator were collected
either by settling on the plates or on the 8 in. x 10 in. (20.3
cm x 25.4 cm) filter downstream from the plates.
Upon completion of each test, the plates and filter were
washed separately with tap water. Samples from each wash were
centrifuged to remove debris, and the masses of methylene blue
collected on the plates and on the filter were determined by
absorption spectroscopy.
The velocity distribution through the settling chamber was
studied to determine the configuration necessary to obtain uni-
form flow. The use of two filters and an extended flared inlet
was necessary to provide the desired velocity distribution.
Figure 6 is a velocity profile measured using a thermal anemom-
eter immediately upstream from the blower before the plates were
in position. From these data, it was evident that the blower
was pulling uniformly across the rectangular opening where the
20.3 cm x 25.4 cm filter was positioned. Figure 7 shows the
velocity profile measured upstream of the collector plates.
u
•z.
HI
o
LLJ
O
UJ
O
O
RECTANGULAR
CIRCULAR
1.
1.0 2.0 4.0 6.0 8.0 10 20 40 60 80 100
GEOMETRIC MEAN DIAMETER, micrometers 4181-126
Figure 4. Theoretical collection efficiency by particle settling in rectangular
and circular tubes.
323
-------
U>
ro
*»
COMPRESSED AIR LINE
REGULATOR
AND TRAP
REGULATOR
DRYER
FLARED INLET
no.
ABSOLUTE FILTER
REGULATOR
VALVE
VIBRATING
AEROSOL
GENERATOR
ROTAMETERS_D.SPERSIpNAIR
I
DISPERSION AIR
FILTER
SETTLING CHAMBER
VARIABLE
VOLTAGE
TRANSFORMER
HIGH VOLUME
AIR SAMPLER
I
SYRINGE PUMP WATER MANOMETER
FUNCTION GENERATOR
4181-74
Figure 5. Apparatus used to measure the collection efficiency of the settling chamber.
-------
The velocity profiles shown in Figures 6 and 7 were con-
sidered satisfactory, and the overall average velocity was used
to calculate the theoretical performance curves for the system.
The average velocity through the chamber, measured using the
thermal anemometer, was in agreement with previous calibration
of the blower.
/
V (m/sec)
0.30
0.20
0.10
STATISTICAL DATA:
MEAN, x , = 0.24 m/sec
STANDARD DEVIATION,
Sx = 0.02 m/sec
COEFFICIENT OF VARIATION,
SX/JT = 8.2%
4181-107
Figure 6. Profile of the air velocity immediately upstream from the
blower of the sett/ing chamber before the plates were positioned.
325
-------
STATISTICAL DATA:
MEAN, x , = 0.23 m/sec
STANDARD DEVIATION,
Sx = 0.02 m/sec
COEFFICIENT OF VARIATION,
Sx/x = 8.8%
4181-108
Figure 7. Profile of the air velocity upstream from the plates
of the settling chamber.
Results—
Two sets of experimental data were obtained from the testing
of the horizontal elutriator. The first set was acquired while
operating the settling chamber at an average gas velocity of
70 cm/sec and a plate length of 38.1 cm. The second data set
was obtained by operating the system at 40 cm/sec after shorten-
ing the plate length to 20 cm. The theoretical curves of the
326
-------
collection efficiency versus aerodynamic particle diameter shown
in Figures 8 and 9 were developed from Equation 5 using the
following parameter values:
Figure 8
p = 1.35 g/cm3
g = 9.8 m/sec2
L = 38.1 cm
y = 181 micropoise
h = 0.701 cm
V = 70 cm/sec
Reynolds No. = 315
Figure 9
p = 1.35 g/cm3
g = 9.8 m/sec2
L = 20 cm
p = 181 micropoise
h = 0.701 cm
V = 40 cm/sec
Reynolds No. = 180
The calculated values from theory for the collection effi-
ciency were found to be in excellent agreement with the measured
values. No corrections were made for end effects.
u
z
UJ
UJ
O
U
2 34 6 8 10 20 30 40
AERODYNAMIC PARTICLE DIAMETER, Mm
60 80 100
4181-109
Figure 8. Theoretical and experimental collection efficiencies for a horizontal elutriator
with rectangular cross-section, plate length 38.1 cm, average gas velocity 70 cm/sec.
327
-------
Collector Design—
The experiments described above indicated that the theo-
retical equations can be used to predict particle collection
by horizontal elutriators to a high degree of accuracy. Con-
sequently, the equations were used to create design nomographs
for inhalable particulate precollectors to be used in conjunction
with the three systems of interest. Design parameters were
calculated for horizontal elutriators to be used with: (1) cas-
cade impactors operated at 14.2 Jl/min and 149°C, (2) SASS
trains operated at 185 £/min and 204°C, and (3) FAST trains
operated at 5,282 Jl/min and 23°C.
Figures 10, 11, and 12 are the design nomographs for the
three trains. On the vertical axis is the open area required,
neglecting the thickness of the tube walls. The horizontal axis
is the length of the elutriator required to yield the inhal-
able particulate performance at the specified flow rate and tem-
perature. In constructing the graphs, it is assumed that the
99
98
95
90
a?
- 80
u
g 70
o 60
u.
Si 5°
2 40
O
P 30
U
3 20
_l
o
° 10
5
2
1
" ^"™
THEO
EXPE
^_>
*"~
RY
RIME
X
NT
/
^
y
/
I /
L/
\
/ /
y
>
/
A
//
^
*
'
^
/
/
^
1 2 3 4 6 8 10 20 40 60 80 100
AERODYNAMIC PARTICLE DIAMETER, Aim 4181-75
Figure 9. Theoretical and experimental collection efficiencies for a horizontal elutriator
with rectangular cross-section, plate length 20 cm, average gas velocity 40 cm/sec.
328
-------
10
VC
RECTANGULAR ELUTRIATOR
2 3 45678
PLATE SEPARATION, mm
| CYLINDRICAL ELUTRIATOR
RECTANGULAR ELUTRIATOR
CYLINDRICAL ELUTRIATOR
TUBE DIAMETER, mm
1.0
2.0
3,0 4.0 5.0 6 7 8 9 10
LENGTH, cm
30 40
50 60 70
4181-124
Figure 10. Relationship of design parameters for horizontal elutriators with
DSQ cutpoints of 15 fim aerodynamic diameter used as precollectors
for instack cascade impactors.
-------
rectangular channels have widths much greater than their heights,
so that the vertical walls will not have a significant effect
on the gas flow.
It can be seen from these graphs that circular tubes can
be used to construct a system having the required performance
with much smaller overall dimensions than rectangular tubes or
channels.
ij|:: :::g|:::::J:::::: :
HJJ5M
i te : lijfEEE^iE; •;::;:
|||p RECTANGU
200i l|^|;i;i||y:]
i: j f : ::::::!i :i±:i:::
~* ^^^H
0 rr ' i^iit" ""It" i
^ i i j
£ 100 U :^::4g:::::::3
< 90' t;!£|JEEEEl|E:i|ijij
< so; ;EE-EE:EEEE;|:;;;;;:
z iiMM4l±iMa|MiJM
0 70 tt: ::EE:|::::i?:::::fi
a 6oJ;;=™=™|:;|;::;
g so - iilHiHHHjNili
OC "T-ff
£ Ez»!»!:;CYL^
:3iii:ii RECT;
30, f := = = = := = = = :::: = :::::::
20 ;; :EEg;:;;:;;:;;;::::
f itt' ±S *--•+-•!
3.0 4.0 5.C
!:!!: ::::!:|:::::!::::::"^
iftfitmi^^^ffl
LAR ELUTRIATOR =
• 4WH5Jg6i 7
ti ||E;:!f!;i|;l;!g^
t X *L i I tf ' V
' t 4 i ' S^
"H! ii-'-i-i--'!-'!
•-::: j iij :|:! i|i:: f ::.
L ' 1 • !-• —1
DRICAL ELUTRIATC
ANGULAR ELUTRIA1
^
) 6.0 8.0 10
iitH-H lll'IIHiUHI I11 ''II!!|1M:
± — |i--± ±— -
8 --: PLATE SEPARATION, m
--ss-,--Sj---oT!«S'"':f • "t —
3, sJj""V«i '!«;''i; ;~r rr* """T
S^---k*k--->sjI!«^1.!ij . ft --J-H
], ^"^ S|-i . -- -i|i.. ,i ,. ^
*i _!,!|_...- j-|j 1 |- l^k -.
f=:i?^5.::|:!s;!j:::*ij 'i J:!|L
>R ;EEEE?s||EE|;g;:!!iii: :: :^|-
-QR Uiiiilp^
DIAMETER, mm :;l; 2'\\ ;EEEE
zEJEEEEEiEEE^::: ;; CYLINDRICA
20 30
LENGTH, cm
jttri i[[i|i jii i [ill I|DJI in ipi
^E^:Mj|j|K|T|||t|yi|l|^
j::||}|l||:;:J
- i re i|"i" "
l:!i!::::!ij! ! ' :: ::: ::±: ::3 :
Si|!iE!!i!ii; I ii; Sj=ji!!!Eii!!!i ii
E!!;!ii±'!!' \ >-«|i:::;;;^: \
Ssl::^!^,::1 i:: j 5|t; !s|l; || :
3 ;-E|;;:4 ;| 5^^6::- 7 J
L ELUTRIATOR i!:::^: i!
40 50 60 80 100
4181-123
Figure 11. Relationship of design parameters for horizontal elutriators with
050 outpoints of 15 jum aerodynamic diameter to be used with
SASS trains.
330
-------
RECTANGULAR ELUTRIATOR
2 3 45678 PLATE SEPARATION, mm
600
500
cee
ECTIONAL AREA, cm2
CO
co
CO
0
CC
o
100
90
80
70
60
50
40
1
V
EEEj=EEEEEE;!!i;i;E:;;;; S iEp!
— ::_::::: :::::::::: - ! : : j : :
= EEEE^EEE-EEEEEE^;E;;!! |!;j^;;: ; EEE!IJ?
E— EE;E^EEEEEE;;;E;E;;;;;;!! ij1 : EEEEE;EE
|;;;... CYLINDRICAL ELUTP
^HiiDiL RECTANGULAR ELUT
1 1""^
20 30
NijISijiiijii; !hSji{l!iJ!!!jipll
ij'iE:;;;;;:1! i E?j=^E:i:!j::!i:;E:Hq
;!ii:!j;;;:; :! ' EEE^I EJE!!!M: !ii 5*2
M
IATOR iTO^Hl|||| lllHTJIcLn
RiATOR: ::EE:i|!i; :::; : ::: :^^
- — . ! k . ' , .
•II M (M
- I .. I ... _y3
-••--• — S
EEE|M;EE;h; :|: EEEEEE TUBE DIAIV
40 50 60 80 100
»
-1
s
s
L
^
IE
-^^-rSr^-
p;
^ .--SN V !,!l... I Sj-,---
NTVJ n*v!\J rnJiJ TrUjTU lirCI JrXJn
~^LJ i iji'j — 'i. i" • i^nJ
,_ rt 5,1 S'ij i 'i ' , r»;
ETER, mm fjjjfl 2^^W3 m
-EE^EEE^E;EEEE -EE;;;;;; CYLIND
200 300
4 — w... ±-i--|-T
i
!'i 'r i"" j t lit t"
^ICAL ELJTRIATOR j- :||
400 500 700 10C
LENGTH, cm
4181-125
Figure 12. Relationship of design parameters for horizontal elutriators with
DSQ outpoints of 15 tim aerodynamic diameter to be used with
FAST.
-------
High-Volume Impactor
The theory of particle collection by impaction has been
developed to a state in which the efficiency can be reasonably
well predicted for ideal conditions if the geometry of the system
is known.9'10 Critical dimensions and parameters of the system
are the jet width, the length of the jet, the jet-plate spacing,
the jet velocity, the Reynolds number of the jet, and the type
of particle collection surface. In addition to being a well-
understood system of particle separation and sizing, impactors
are very compact in size.
However, impactors can suffer from a severe limitation if
the particles or substrates are not adhesive. Particles much
larger than the D50 have greater stopping distances and may
strike the substrates with high velocities and rebound. This
phenomenon results in an efficiency curve that rises with par-
ticle diameter to a maximum, frequently at less than 70-80 per-
cent, and then declines to much lower efficiencies for still
larger particles.1'2'3
Experimental Procedures—
A prototype impactor was constructed using six first stages
from a commercially available cascade impactor (Model 35, Sierra
Instruments) in parallel. The impactor is designed to operate
at a flow rate of about 0.020 m3/sec. In this study, which was
performed before the inhalable particulate performance curve
was developed, the flow rate through each impactor was approxi-
mately 0.015 m3/sec. The reduced flow rate was selected to
achieve 95 percent collection of 15-um particles. The impactor
stages were covered either with glass fiber substrates or grease.
The FAST including the prototype impactor stage is shown
schematically in Figure 13. The calibration aerosols were gene-
rated using a vibrating orifice aerosol generator as described
above. For these tests, however, ammonium fluorescein particles
were used and the cleanup procedure was slightly different.
To help ensure the recovery of the aerosol particles trapped
by the grease (petroleum jelly) on the impactor collection plates,
they were washed with benzene in an ultrasonic cleaner in addi-
tion to being washed with ammonium hydroxide. Glass fiber sub-
strates were washed with ammonium hydroxide.
As before, the mass of material collected on each surface
was determined by absorption spectroscopy.
Results—
The results of a limited calibration study are shown in
Figure 14. It can be seen that particle bounce is severe with
332
-------
INLET
THERMOCOUPLE
CYCLONE
FILTER
INLET
CYCLONE
TRAP
OUTLET
THERMOCOUPLE
VACUUM
PUMP
EXHAUST
4181-95
Figure 13. Fugitive air sampling train (FAST) components.
both glass fiber and greased substrates. Because of these un-
satisfactory results, the experiments with impactors were aban-
doned in favor of the horizontal elutriator. It should be noted
that stages 2-5 of the same type of impactor have been calibrated
by Willeke and shown to behave well with liquid particles.11
Cyclones
Although the flow in cyclones is more complicated than
in impactors and no theory exists to adequately predict their
behavior, several experimental investigations have been performed
demonstrating their utility for separating and sizing small par-
ticles. 1 2 ': 3'x ** Smith et al.,15 and John et al.,16 have demon-
strated that the curves of particle collection efficiency for
small cyclones can be as steep as those of impactors, and signi-
ficant reentrainment does not occur. Chan and Lippmann17 have
shown that experimental efficiency data for small cyclones can
be fitted well using the empirical relationship:
Eff = 0.5 + 0.5 tanh
+ A-2B
(6)
where
Q is the sample flow rate, cm3/sec,
333
-------
D is the diameter of the sampled particle, cm, and
A,B,K,n are empirical constants.
Also it was shown that the D50 vs. flow rate relationship
is given over a limited range of sampling conditions by:
D50 = KQ
n
(7)
100
90
80
7°
o
ui 60
z 50
o
o
u 40
O
O
30
20
10
GREASED PLATES O
GLASS FIBER SUBSTRATES •
.
X
1
/
/
f
1
i
i
f
O
'-«N
\ o
\
\
\
\
u
1
I
1
1
1
•1
2 3456 810
AERODYNAMIC PARTICLE DIAMETER, jum
20
4100-13
Figure 14. Particle collection efficiency for the impactor.
334
-------
From their own research and data reported by others, Chan
and Lippmann reported values of K ranging from 6.17 to 4591,
and values of n from -0.636 to -2.13. Smith et al. reported
values of K from 14.0 to 44.6, and n from -0.63 to -1.11. Al-
though the trend in the data is for n to have larger values for
small cyclones (D50's) and K to be larger, the correlation is
not consistent enough to be predicted from the cyclone geometry,
and no data are reported for temperatures other than ambient.
Furthermore, there are definite discontinuities and hysteresis
effects in the relationship given in Equation 7, even for indivi-
dual cyclones, as the flow is increased and decreased. The dis-
continuous and hysteresis effects are generally attributed to
transitions from turbulent to laminar flow, or the reverse, in
the outlet tube of the cyclone as the flow is changed.13
Smith et al. found the D50 of cyclones to increase linearly
as the gas temperature and viscosity were increased; but again
the rate of increase was not predictable from the cyclone geometry.
Certainly a modification to Equation 7 would be required to pre-
dict any temperature dependence. Lacking an adequate theory
for predicting the performance of cyclones before they are de-
signed and calibrated, it was found necessary in this study,
as in previous work, to extrapolate the dimensions for a new
cyclone to give the desired performance from those of similar
cyclones of known performance.
However, because of these difficulties, and the requirement
that the inhalable particulate cyclones always be operated to
yield D50's at 15 ± 2 urn, it was still considered necessary to
calibrate them over a range of temperatures and flow rates similar
to those expected in field operation.
Experimental Procedures—
The objective of this phase of the program to develop sam-
plers for inhalable particles was to design and evaluate two
systems: (1) a cyclone to be used as the precollector for cas-
cade impactors, and (2) a system, consisting of two cyclones
and a filter in series, to be used as the primary system for
measuring the total particulate concentration, the inhalable
particulate concentration, and the fine particle concentration.
Both systems are to be used in process streams where the particle
concentration and temperature are generally much higher than
in the atmosphere. The precollector for impactors and the first
cyclone in the series train must have efficiency curves satis-
fying the specifications for inhalable particulate samplers shown
in Figure 4. The second cyclone in the series train must be
designed to have a D50 of 2.5 ± 0.5 ym aerodynamic.
The nominal operating conditions used in designing the
systems were flow rates of 14 &/min and 21 Jl/min at 150°C,
respectively, for the precollector and cyclone systems. With
335
-------
these operating conditions as a goal, the dimensions of the
cyclones were extrapolated from those previously evaluated.15
Figures 15 and 16 are schematic illustrations of the two
new systems. The precollector cyclone is designated SRI-IX and
the large cyclone in the series train, SRI-X. Cyclone SRI-III,
which had been previously evaluated as part of the SRI/EPA 5-
stage cyclone train, was found to be adequate for the smaller
cyclone in the new system.
The critical dimensions of all three cyclones are given
in Figure 17.
In order to calibrate the cyclones at elevated temperatures,
the heating arrangement shown in Figure 18 was used in conjunc-
tion with the vibrating orifice aerosol generator described above.
Tests were made at temperatures of 23°C, 93°C, and 150°C. Am-
monium fluorescein particles were used for calibrating the cyc-
lones. Samples were taken frequently of the heated aerosol to
ensure that the calibration system was stable and that the par-
ticles were spherical and of the proper size.
In this study, the primary objective was to determine the
operating conditions under which the cyclones satisfied the de-
sign criteria. For this purpose, it was sufficient to use an
abbreviated calibration procedure and thus to reduce the exten-
sive amount of testing that would be needed for complete calibra-
tion. At a specific temperature, monodisperse particles having
nominal aerodynamic diameters of 15 urn were generated and sam-
pled. Tests were made at a variety of flow rates to determine
the sampling rate required to yield a D50 of 15 ym at the given
temperature. Only limited data were taken to determine collec-
tion efficiency vs. particle diameter.
SAMPLING NOZZLE
CYCLONE
D50 = 15 ± 2 ;um
CASCADE IMPACTOR
PROBE
4181-76
Figure 15. Schematic of a cascade impactor/precollector cyclone system.
336
-------
When preparing to generate monodisperse particles using
the vibrating orifice method, it is necessary to know the solu-
tion flow rate and frequency of vibration in order to calculate
the concentration of solute required to yield the desired par-
ticle diameter after drying. In practice, the flow rate can
be selected and the solution prepared precisely, but the fre-
quency at which the generator is finally found to yield maximum
stability is unpredictable to some extent. Therefore, as indi-
cated in the figure captions, the actual particle diameters dif-
fered slightly from 15 urn. In these tests the particles had
aerodynamic diameters of 15.0 ± 0.6 jam.
Results—
Figures 19-22 show the calibration data for the three cyclones,
In Figure 19, the particle collection efficiency is plotted vs.
flow rate for cyclone SRI-IX.
As indicated in the figure, data were taken at 23°C, 93°C,
and 150°C. Similar data are shown in Figure 20 for cyclone SRI-X.
The settling velocity in air of 15 ym particles is 0.7 cm/sec
and there was some concern that settling might influence the
collection efficiency of the larger cyclones by making them sensi-
tive to orientation. In Figure 20, which contains calibration
data for cyclone SRI-X, data are shown taken with the cyclone
in both vertical and horizontal orientations. It can be seen
that there is little, if any, effect due to particle settling.
SAMPLING NOZZLE
PROBE
CYCLONE SRI-IX
D50 = 15 +2 f*"1
CYCLONE SRI III
D50 = 2.5 ± 0.6 Hm
4181-120
Figure 16. Schematic of two-cyclone system.
337
-------
T
i
-H B
1cup
D
cup"
DIMENSIONS (CENTIMETERS)
CYCLONE
SRI-X
SRI-HI
SRI-IX
D
6.14
3.11
5.12
Din
1.83
0.75
1.53
De
2.17
0.83
1.81
B
2.92
0.76
2.43
H
8.47
4.91
7.06
h
2.82
1.40
2.35
Z
5.65
3.51
4.71
S
2.40
1.08
2.00
Hcup
2.635
2.22
2.26
Dcup
6.14
3.10
5.12
4181-22
Figure 17. Summary of cyclone dimensions.
338
-------
AEROSOL STREAM
FROM VIBRATING
ORIFICE AEROSOL
GENERATOR
IVBSOLUTE FILTER
OVEN
, KEPT AT
I AEROSOL TEMPERATURE
HEAT I
EXCHANGER
SAMPLING
PORT
MERCURY WATER
MANOMETER MANOMETER
4181-92
Figure 18. Calibration system for heated aerosols.
The data reported in Figures 19 and 20 will be used to
select the flow rates of the new sampling trains. Since there
is no adequate theory for calculating cyclone efficiency, cyclone
SRI-III was calibrated at the same flow rates required for proper
operation of cyclone SRI-X. These calibration data are shown
in Figure 21. D50's and flow conditions for the three cyclones,
as derived from the graphs, are:
Cyclone SRI-III
3.1 vim (aerodynamic) D50 at 23°C, 11 S,/min
2.6 urn (aerodynamic) D50 at 93°C, 20 Jl/min
2.3 ym (aerodynamic) D50 at 150°C, 23 &/min
Cyclone SRI-IX
15 ym (aerodynamic) D50 at 23°C, 6.8 5,/min
15 ym (aerodynamic) D50 at 93°C, 12 £/min
15 ym (aerodynamic) D50 at 150°C, 14 &/min
339
-------
u>
^
o
!
.<•£>
O C~5
~s O
Q) 5'
Cb Ch'
II
fb
3
^
COLLECTION EFFICIENCY,
S 8
y> o
~j
CO
§
Ul
T-
3
oT
n>
o
a
*3
H-
O
0
in
o
(a
ft
,_,
Ul
o
o
O
INJ
CO
J°
\
3
H-
3
en
-c
3
oT
fD
O
*3
(a
3
H-
O
o
Ul
o
JU
ft
VD
co
o
O
N>
0
X.
3
H-
3
Ul
t
3
aT
n>
O
a
3
(u
3
J— «.
O
D
Ul
0
a>
ft
fO
CO
O
O
H1
I-1
\
3
H-
3
O
<
O
M
O
3
n>
en
-------
Also, limited data were taken with particles with diameters
other than 15 ym using cyclone IX. These are shown plotted with
the inhalable particulate performance specifications in Figure 22.
The data reported above define the operating points of
cyclones SRI-IX and SRI-X at three discrete values of tempera-
ture. Figure 23 shows the data plotted on semi-log coordinates
with a smooth line fitted to the points. The equations of flow
rate vs. viscosity are:
For cyclone SRI-X, Q = (105 logy - 225H/min
and for cyclone SRI-IX, Q = (69.5 logy - 150H/min.
It is not known at present how accurately these equations
can be extrapolated to temperatures greater than 150°C.
100
80
o
HI
O
JZ 60
IL
111
z
o
o
o
40
20
1 I 1 1 1 1 1
0 VERTICAL, OUTLET UP
£ VERTICAL, OUTLET DOWN
• A H
IORI2
!ON1
FAL
/
_j
/
/
/23°C
/
93°C /
//
//
/
/
« it
J
/
r
A
150°C
4 6 8 10 20 40
FLOW RATE, £/min 4181-7S
Figure 20. Collection efficiency of cyclone X vs. flow rate for particles of
15±0.6 urn aerodynamic diameter.
341
-------
In order to gain a better qualitative understanding of the
mechanisms governing the performance of cyclones, Table 1 was
prepared summarizing some of the dimensions and operating param-
eters of cyclones evaluated at Southern Research Institute.
The parameters of greatest interest are the Stokes number of
15 pm particles in the inlet and the ratio of the settling ve-
locity to the centripetal velocity of 15 ym particles in the
cyclone body. The latter is calculated assuming that the air
flow in the cyclone body is a jet having the same velocity and
diameter as in the inlet. This is obviously a crude approxima-
tion, but probably accurate enough to allow a reasonable esti-
mate of the desired ratio.
as
u
01
O
01
Z
O
u
01
_i
O
u
AERODYNAMIC PARTICLE DIAMETER, jLtm
4181-87
Figure 21. Collection efficiency vs. aerodynamic particle diameter for
Cyclone III at 22° and 11.3 Wmin (Q), 93°C and 19.8 l/min
(Q),and 150°C and 22.7 Z/min
342
-------
The settling velocity Vs of unit density spheres is given
by the expression (equivalent to Equation 2):
Vs = mgB
where
m is the particle mass,
3 is the particle mobility, and
g is the acceleration due to gravity,
The centripetal velocity VG is given by:
\7 -
VC ~
mV.
ln
R
V
in is the gas velocity in the inlet and
where
R is the cyclone radius.
The desired ratio, then, is:
V
s
(8)
(9)
Vc
99
98
95
90
0 80
Z
UJ
0 70
£ 60
Z 50
O a.n
COLLECTI
-» K> CO J
-» co en o o o c
V. 2
in
/
/
/
f
/
,
./
-------
30
25
20
^5
oc
o
_1
LL
10
150
CYCLONE X
Q = 105 log^ -225
CYCLONE IX
Q = 69.5 log/u -150
I
I
200 250 300
VISCOSITY (ju), micropoise
400
4181-121
Figure 23. Gas flow rate versus viscosity at D^Q = 15
diameter for cyclones IX andX.
aerodynamic
344
-------
TABLE 1, OPERATING PARAMETERS OF SRI CYCLONES
u>
Cyclone
Number
V
IV
III
II
I
BRINK
IX
X
Q (cm3/sec)
472
472
472
472
472
28
233(150°C)
190
D(cm)
1.52
2.54
3.11
3.66
4.47
1.27
5.11
6.14
D.
in
0.
0.
0.
1.
1.
0.
1.
1.
(cm)
30
57
75
01
27
51
53
83
Vin (cm/sec)
6670
2300
1100
600
370
130
130
70
Ap(mm H20)
1800
330
70
40
5
-
-
-
DSO
0.
0.
1.
2.
5.
13
15
15
(pm)
32
65
4
1
4
/St
0.1
0.1
0.1
0.1
0.2
0.6
0.3
0.2
gR/vin2
i.
2.
1.
5.
1.
4.
6 x 10"5
3 x 1C""
3 x 10"3
0 x 10~3
6 x 10"2
0 x 10~2
0.2
0.6
Q is the sample flow rate
D is the diameter of the cyclone body
Din is the diameter of the inlet
Vj[n is the gas velocity in the inlet
AP is the pressure drop across the cyclone
St is the Stokes number of the 15-ym particles in the inlet
gR/Vj_n is the ratio of the settling velocity to the centripetal velocity in the cyclone
body
-------
The values of /St and VS/VC can be used to estimate whether
or not impaction onto the wall opposite the inlet or particle
settling is likely to play a_large part in the collection ef-
ficiency of a cyclone. If /St is less than about 0.4, then
impaction is not likely to occur.10 If VS/VC is very small,
then settling is not likely to occur. From Table 1, it can be
seen that impaction is probably very significant in the cyclone
used with the Brink impactor, while settling probably contri-
butes somewhat to the collection efficiency of cyclones SRI-IX
and SRI-X.
SUMMARY AND CONCLUSIONS
Three systems have been developed and evaluated for the
purpose of sampling inhalable particles without reentrainment:
(1) a horizontal elutriator for sampling fugitive and ambient
aerosols, (2) a cyclone precollector to be used in-stack with
cascade impactors, and (3) a series-cyclone train for in-stack
use. Each of the systems was calibrated under typical sampling
conditions and found to perform within the range specified for
samplers of inhalable particles. A fourth method, a high volume
impactor, was tested and rejected because of poor performance
due to particle bounce.
Second generation versions of the systems have been con-
structed for use in the field to determine emission factors for
inhalable particles from stationary sources.
ACKNOWLEDGEMENT
This research was supported by the U.S. Environmental Pro-
tection Agency under contract 68-02-3113.
REFERENCES
1. Corn, M., and F. Stein. Re-entrainment of Particles from
a Plane Surface. Am. Ind. Hyg. Assoc. J. 26:325-336, 1965.
2. Rao, A.K., and K.T. Whitby. Non-ideal Collection Character-
istics of Single Stage Cascade Impactors. Am. Ind. Hyg.
Assoc. J. 38:174-179, 1977.
3. Gushing, K.M., J.D. McCain, and W.B. Smith. Experimental
Determination of Sizing Parameters and Wall Losses of Five
Source-Test Cascade Impactors. Environ. Sci. Technol.,
13:726-731, 1979.
4. Fuchs, N.A. The Mechanics of Aerosols. Pergamon Press,
New York, 1964.
5. Stein, F., W.A. Esmen, and M. Corn. The Shape of Atmospheric
Particles in Pittsburgh Air. Atmos. Environ. 3:443-453,
1969.
346
-------
6. Rimberg, D., and J.W. Thomas. Comparison of Particle Size
of Latex Aerosols by Optical and Gravity Settling Methods.
J. Colloid Interface Sci. 32:101-105, 1970.
7. Corn, M., F. Stein, Y. Harnmad, S. Manekshaw, W. Bell, S.J.
Penkola, and R. Freedman. Physical and Chemical Character-
istics of "Respirable" Coal Mine Dust. Ann. N.Y. Acad.
Sci. 200:17-30, 1972.
8. Mercer, T.T. Aerosol Technology in Hazard Evaluation.
Academic Press, New York, 1973. pp. 192-200.
9. Ranz, W.D., and J.B. Wang. Impaction of Dust and Smoke
Particles. Ind. Eng. Chem. 44 (6):1371-1380, 1952.
10. Marple, V.A. A Fundamental Study of Inertial Impactors.
Ph.D. Thesis, University of Minnesota, Mechanical Engineer-
ing Department, Minneapolis. Particle Technology Laboratory
Publication 144, 1970. 243 pp.
11. Willeke, K. Performance of the Slotted Impactor. Am. Ind.
Hyg. Assoc. J. 36:683-691, 1975.
12. Rusanov, A.A. Determination of the Basic Properties of
Dust and Gases. In: Ochistka Dymovykh Gasov v Promyshlennoy
Energetike [Cleaning Stack Gases in Industrial Power Engineer-
ing] , A.A. Rusanov, I.I. Urbakh, and A.P. Anastasiadi, eds.
"Energiya," Moscow, 1969.
13. Hochstrasser, J.M. The Investigation and Development of
Cyclone Dust Collector Theories for Application to Miniature
Cyclone Presamplers. Ph.D. Thesis, University of Cincinnati,
Cincinnati, 1976.
14. Lippmann, M., and T.L. Chan. Cyclone Sampler Performance.
Staub Reinhalt. Luft 39:7-12, 1979.
15. Smith, W.B., R.R. Wilson, and D.B. Harris. A Five-Stage
Cyclone System for In-Situ Sampling. Environ. Sci. Technol.
13(11):1387-1392, 1979.
16. John, W.G., P. Reischl, and J.J. Wesolowski. Size-Selective
Monitoring Techniques for Particulate Matter in California
Air. AIHL/SP-12. Final Report, California Air Resources
Board Contract No. A5-00487, 1978.
17. Chan, T.L., and M. Lippmann. Particle Collection Efficien-
cies of Air Sampling Cyclones: An Empirical Theory. Environ.
Sci. Technol. 11:377-382, 1977.
347
-------
PAPER 19
DEPOSITION OF INHALED PARTICLES AND POSSIBLE
SAMPLING METHODS
VITTORIO PRODI
GIUSEPPE TARRONI
CARLO MELANDRI
LABORATORIO DI FISICA SANITARIA
DIPARTIMENTO RADIAZIONI, CNEN
ABSTRACT
Inhalation is one of the most important routes of internal
contamination. The efficiency of incorporation depends on aerosol
properties and on fluid dynamic properties of the airways. In
this paper total and regional deposition in humans will be briefly
reviewed with the aim of understanding the basic mechanisms re-
sponsible for particle deposition and assessing the sensitivity
of deposition to the main aerosol, physiological, and possibly
pathological factors. The available data on sampling efficiency
of the airways will be also reported in order to reach a detailed
picture of the overall transfer efficiencies.
Two approaches to aerosol characterization for inhalation
toxicology will then be presented. One is based on a direct
size distribution measurement with a recently proposed aerosol
spectrometer and the other is based on the simulation of the
regional deposition. The first can be referred to the cut-off
sizes that are being introduced by EPA, while the second, if
a particular breathing pattern and individual deposition are
assumed, can yield directly the amount deposited in each region.
INTRODUCTION
The atmospheric environment may contain airborne contami-
nants that pose serious risks of incorporation and then of toxic
action. Particulate contamination is of great concern and is
of challenging difficulty to a quantitative assessment of in-
halation risks.
Incorporation of particulate matter is a step process that
comprises inhalation of the aerosol, deposition of the inhaled
aerosol, and clearance or translocation of the deposited aerosol,
and the final result is the retention of the contamination with
the resulting toxic action.
348
-------
Each one of these steps has measured or measurable effi-
ciencies that depend on aerosol properties, on airways configura-
tion, and on respiration patterns. Very detailed reviews have
been published by the ICRP Task Group1 on Lung Dynamics and by
Lippmann.2 Therefore here only the most recent results will
be reported, with emphasis on the parameters relevant to the
risk of incorporation.
Definitions
Inhalability: is the probability for an airborne particle to
enter the airways.3
Deposition: is the probability for an inhaled particle to
touch a surface of the respiratory tract and to
adhere to it.3
Total deposition: is such a probability referred to the entire
respiratory tract.
If a more detailed picture of the contamination
is needed, the respiratory tract is divided into
regions, characterized by predominant deposition
and, for slowly soluble particles, clearance me-
chanisms. There is a generally accepted defini-
tion of the regions:1'
Extrathoracic airways, in which the deposition is mainly due
to inertia and particles are cleared within min-
utes either by mechanical transport of particles
or secretions.
Tracheobronchial airways, in which particles are deposited by
inertia and settling and from which particles
are removed within hours by mechanical transport
of secretions.
Alveolar air spaces, characterized by small particle-to-wall
distances, in which particles are deposited mainly
by gravitational settling and brownian diffusion.
The removal takes months or even years, mainly
by phagocytosis, and subsequent cell transport
to the mucociliary escalator or to the lymphatic
system, or by dissolution.
Regional deposition is the probability for an inhaled particle
to reach a surface of the given region and adhere
to it. Total deposition is the sum of regional
depositions.
349
-------
Retention: total retention is the probability for a deposited
particle to be retained in the body, while regional
retention is such a probability referred to a
given region of the respiratory tract.
Deposition mechanisms: particles are transported across stream-
lines by a number of mechanisms: if they touch
a surface, they deposit there. The main mechanisms
are gravitational settling, brownian diffusion,
inertial impaction, and electrostatic attraction
due to image forces. Experimental data are avail-
able for spherical or compact particles and here
interception will not be considered. All of these
mechanisms are strongly size dependent. Where
aerodynamic effects prevail (settling and impac-
tion) , the
aerodynamic diameter, of a particle, defined as the diameter
of a 1 g cm density sphere having the same settl-
ing velocity as the particle, is a satisfactory
and unique parameter accounting for deposition
and aerodynamic techniques should be preferred
for characterization. Where diffusion prevails,
since the diffusion coefficient is independent
of density, the aerodynamic diameter is not repre-
sentative and a
diffusive diameter should be introduced. ** This could be defined
as the diameter of the sphere that has the same
diffusive properties as the unknown particle.
It is therefore important that characterization
of an aerosol in this range should be based on
diffusion or on techniques that can be uniquely
related to diffusion. In the case of very large
densities the range of overlap of diffusion and
aerodynamic effects varies and one can still have
aerodynamic deposition with appreciable diffu-
sive deposition.
AIR FLOW IN THE RESPIRATORY TRACT
Deposition depends strongly on the flow which is effective
in airways, because it can affect the time-dependent mechanisms
through the mean residence time, MRT, (time dependent mechanisms
are settling and diffusion; electrostatic deposition is also
time-dependent,1"3 at least in the range 0.3 to 0.6 ym) and the
"prompt" mechanisms (impaction) through the volumetric flow
rate,3 VFR. A detailed description of the aerodynamic properties
of the respiratory tract has been given by Lippmann2 and here
will just be briefly reported.
350
-------
Respiratory frequency and tidal volume rule the MRT and
the VFR. Mixing between tidal and residual air has a very im-
portant role in transferring inhaled particles into the residual
volume and thereby strongly increasing their residence time.
Mixing can take place in the conducting airways due to turbulence
and to vortices that even in laminar flow are established in
the bifurcations5 and propagate in the following branches. The
mixing mechanism which is important in the whole respiratory
tract is due to the non-uniformity, both geometric and dynamic,
of the airways. Their diameter in each generation is consider-
ably scattered; branching angles, also, are asymmetric; this
produces a non-uniform flow in the following bifurcations which
is not reversed in the exhalation.
The overall effect is a considerable mixing.6 This has
been studied theoretically and an effective coefficient can be
introduced that accounts for the observed effect.7'8'9
This non-uniformity can be enhanced by the relatively rigid
structures (blood vessels and bronchi) of the airways: during
expansion and contraction geometric similarity is not preserved.
The mixing effect is enhanced at lower values of the functional
residual capacity and this can account for the increased deposi-
tion with decreasing expiratory reserve volume,9 ERV.
Accurate morphometric data are not available for populations;
it is not possible therefore to correlate the scatter of deposi-
tion values to any distribution of morphometric parameters, but
there is little doubt that this scatter should be due to anatomi-
cal as well as physiological or pathological differences.2
TOTAL DEPOSITION
Total deposition has received greater attention than regional
deposition, since it can be studied in vivo with relatively simple
approaches. Two techniques will be considered here: one10'11'12
is based on the measurement of inhalation and exhalation flow
rates and particle concentration just at the entrance of the
airways. The numbers of inhaled and exhaled particles and the
volumes are then computed by multiplication and integration.
The other13 is based on the measurement of the particle
concentration averaged over several cycles with the volunteer
(exhalation) and without the volunteer (inhalation) and on a
separate measurement of respiratory volumes.
The two techniques have been recently compared and gave
coincident results.13
Heyder and coworkers3'x ** have thoroughly investigated total
deposition, DE, as a function of route of inhalation, particle
size, and respiratory parameters.
351
-------
Size
The behavior of DE as a function of size is shown in Fig-
ure 1 (from Heyder et al.3) for mouth breathing. The general
trend shows a minimum around 0.5 ym and increases both for de-
creasing and for increasing size. Below 0.5 ym the increase
is due to increasing diffusion coefficient. Unfortunately in
the ultrafine range experimental data is scarce.15 Above 0.5 ym
deposition is due to gravitation and impaction.
MRT
Figure 1 shows also the effect of MRT on total deposition:3
DE increases with increasing MRT because both diffusion and set-
tling are time dependent. Impaction becomes important at higher
sizes and this is shown by the smaller effect of MRT at high
VFR.
1.0
o
0.8
-.0.6
O
O 0.4-
LU
Q
0.2
MEAN RESIDENCE TIME, 1-8 s
VOLUMETRIC FLOW RATE. 250 cm3/s
PARTICLE DENSITY, 0.91 g/cm3
2468
PARTICLE DIAMETER, jum
10
Figure 1. Effect of particle size and mean residence time on total deposition
(from Heyder et a1.3).
352
-------
VFR
Total deposition is independent of VFR up to 1 ~ 1.5 ym
since impaction is not effective for the range of VFR encountered.
At larger sizes VFR begins playing a more and more important
role, as shown in Figure 2 (from Heyder et al.3). The importance
of impaction is depicted in Figure 3, where MRT and VFR are varied
while keeping the tidal volume (TV) constant. There is a definite
cross-over of the curves, which is even more dramatic for nose
breathing, as shown in Figure 4 (also from Heyder et al.3), where
it takes place around 1 ym, showing the contribution of impac-
tion to nose deposition.
1.0
0.8-
o
1 0.6
•2.
g
H 0.4
CO
O
Q.
UU
a
0.2-
MEAN RESIDENCE TIME, 1 s
VOLUMETRIC FLOW RATE, 250-1000 cm3/s
PARTICLE DENSITY, 0.91 g/crn3
1000
"750
500
250
2468
PARTICLE DIAMETER, jum
10
Figure 2. Effect of particle size and volumetric flow rates on total deposition
(from Heyder et al.3).
353
-------
1.0
MEAN RESIDENCE TIME, 1-8 s ]
VOLUMETRIC FLOW RATE. 125-1000 cm3/s
PARTICLE DENSITY, 0.91 9/cm3
0.8-
o
0.6
O
CL
0.2
_ 1 1000
2 500
4 250
8 125
2468
PARTICLE DIAMETER, //m
10
Figure 3. Effect of particle size on total deposition for mouth breathing
at 1000 c/r>3 tidal volume (from Heyder et al.3).
1.0
'MEAN RESIDENCE TIME, V8 s
VOLUMETRIC FLOW RATE, 125-1000 cm3/s
PARTICLE DENSITY, 0.91 g/cm3
0.8-
o
*:
u
CD
£ 0.6
55 0.4
O
a.
LU
a
0.2
1 1000
2 500
4 250
8 125
2468
PARTICLE DIAMETER, ptm
10
Figure 4. Effect of particle size on total deposition for nose breathing at
1000 cm3 tidal volume (from Heyder et al.3).
354
-------
Biological Variability
It is now generally accepted2 that even under strictly con-
trolled breathing conditions and residual volumes there is a
definite intersubject variability of total deposition. An ex-
ample of this is given16 in Figure 5, where the DE is plotted
as a function of particle size between 0.3 and 1.5 ym unit density
spheres for six volunteers, breathing at 1000 cm3 TV and 15
respiration/min, each at his own expiratory reserve volume,
ERV. This is interesting since in this range total deposition
is also alveolar deposition.
The scatter of data reaches a factor of 2 and cannot be
explained on the basis of respiratory parameters.
For each volunteer instead, with 0.6 pro aerosols, a marked
dependence on ERV is found. The relative DE can be expressed
as a -1/3 power of the ERV relative to normal, probably due to a
stronger mixing with smaller volumes;9 a 10% variation of DE
corresponds to a 30% variation of ERV around the normal value.16
40
30
S20
Q.
111
Q
EXPIRATORY RESERVE
SUBJECT VOLUME, cm3
15 1080
24 1820
6 1210
1730
1400
2190
0 0.5 1.0 1.5 2.0
SPHERE DIAMETER, urn, density 1 g/cm3
Figure 5. Total deposition for a group of 6 volunteers, at 1000 cm3 tidal
volume and 15 respirations/min (from Tarroni et al. 16),
355
-------
Electric Charges
Electric charges carried by aerosol particles have a definite
effect on deposition, leading to an increased efficiency. This
has been found in deposition measurements of aerosols produced
by atomization; when the aerosol was not neutralized, deposition
was higher and poorly reproducible.
Quantitative measurements16'17 have been performed only
with monodisperse particles charged with positive elementary
charges in a narrow number distribution and in controlled breath-
ing conditions. The electrostatic deposition, shown in Figure 6,
increases monotonically with increasing charge number for 0.3
and 0.6 ym unit density spheres and is due to image forces17
and therefore connected with the charge individually carried.
This has been confirmed theoretically18 in the concentra-
tion range considered.
Quantitative information on charge distribution of actual
aerosols is lacking; for aerosols freshly generated by disruption
(grinding and atomization) the absolute charge can be very high
and its contribution to total and regional deposition should
be evaluated.
16
14
u? 12
Q
2 10
O
CO
O
QJ 6
Q
.SUBJECT 9
•SUBJECT 15
0.6 j
0.3
10 20 30 40 50 60
ELECTRON CHARGES, n
70 80
Figure 6. Contribution to total deposition from electric charges carried by
0.3 and 0.6 yjn particles (from Tarroni et at. 16).
356
-------
REGIONAL DEPOSITION
Nose Deposition
Nose deposition must be treated separately since the breath-
ing route can be to some extent a matter of choice for the sub-
ject, and has peculiar clearance pathways; this is important
both for total and regional deposition.
Nose deposition data19'20'12'21'22 have been recently sum-
marized by Lippmann2 and are shown in Figure 7.
The deposition is a linear function of the logarithm of
the inertia parameter D2F when F is the VFR and is practically
quantitative for 9 ym particles at a flow rate of 30 liters per
minute.
The results are in fairly good agreement with the ICRP Task
Group deposition curve, based on Pattle's23 data.
Heyder and Rudolf 21* have also successively studied in detail
the deposition in the nose during exhalation, which is fairly
close to deposition during inhalation, although this efficiency
applies only to the transmitted fraction.
AERODYNAMIC DIAMETER AT 30g/min, M™
1 2 4 6 8 10 20
* 100
1»» O
RutfoM ft Hndv t»M
10,000
Figure 7. Head deposition during inhalation via the nose vs the impaction
parameter D?F (from Lippmann?). The solid line is the
curve based on Pattle's23 data.
357
-------
Mouth Breathing
Extrathoracic Deposition—
Regional deposition in mouth breathing has been studied
by Lippman,2 Lippmann and Albert,25 Chan and Lippmann,26 and
by Stahlhofen et al.27 by external counting of labeled deposited
particles.
Head deposition too can be linearly fitted with the Ig D2F
parameter. Lippmann's2 data have been extrapolated to obtain
the size for quantitative head deposition, which is around 17 urn
for a 30 8- min-1 VFR.
Stahlhofen et al.'s27 data show a slightly higher efficiency
pointing to 100 percent deposition around 11 ym at the same flow
rate.
The inertial parameter is not fully representative of deposi-
tion since the geometry of the airways may be dependent on the
flow.
This should be considered in establishing effective thresh-
olds of sampling instruments and guidelines. In Figure 8 the
average extrathoracic deposition of three subiects27 is reported
for two flow rates together with Lippmann's curve.2
In Figure 9 the effective total and regional depositions
are shown as the average of three subjects, for two breath-
ing patterns: TV = 1500 cm3, 15 resp.min"1 (MTR = 2 sec, VFR
= 750 cm3 sec"1) and TV = 1000 cm3, 7.5 resp.min"1 (MRT = 4 sec,
VFR = 250 cm3 sec"1)- For the extrathoracic deposition the
curves are derived from the same data points of Figure 8.
Tracheobronchial Deposition—
Tracheobronchial deposition has been studied in vivo by
Lippmann, Albert, and Peterson,28 on a large number of volunteers,
ranging from healthy non-smoking to smoking, mild bronchitic
and severe bronchitic. Lately it has been studied by Chan and
Lippmann26 both in hollow casts and in vivo and by Stahlhofen,
Gebhart and Heyder27 on three healthy subjects. In addition,
detailed studies have been performed on hollow casts of human
bronchial trees by Chan, Schreck, and Lippmann,29 that have
pointed out the flow pattern in the trachea, and preferential
deposition sites in connection with air flow and turbulence.
In addition the effect of electric charges on hollow cast deposi-
tion has been examined30 and the dependence on image forces has
been confirmed.
The studies of Chan and Lippmann26 have shown a remarkable
biological variability of deposition data even in the tracheo-
bronchial tree.
358
-------
1.0-
0.8-
0.6-
Volumetric flow/rate, cm3/s 250
Mean residence time, s 4
Subject 1 O
Subject 3 A
Subject 4 D
--- Lippmann (mean)
Flow rate 370 - 680 cm^/s
mean residence time 2.1 s
750
2
103
Figure 8. Head deposition during inhalation via the mouth vs. impaction parameter. The
solid lines and points are for three subjects at two volumetric flow rates (from
Stahlhofen et al.27)r while the broken line is Lippmann's2 fitted curve.
z
o
00
O
a.
HI
0
1.0
0.8
,0.6-
0.4-
0.2 i
Mean residence time, 4s
Volume flow rate, 250
Mean residence time, 2s
Volume flow rate, 750 cm-Vs
TOTAL
EXTRATHORACIC
TRACHEO-
BRONCHIAL
0 2 4 6 8 10
AERODYNAMIC DIAMETER, Mm
Figure 9. Average total and regional depositions for three subjects at two
different respiratory patterns and for mouth breathing (based
on data of Stahlhofen et al.27).
359
-------
Figure 10 shows the TB deposition expressed as a fraction
of the aerosol entering the trachea. The straight lines rep-
resent the average and the scatter of the values found by Lipp-
mann,2 while the data of Stahlhofen et al.27 are shown by the
points. These show a smaller scatter of data and two distinct
behaviors at two rates as well as values of deposition slightly
lower than Lippmann's average.
The actual tracheobronchial deposition is a bell-shaped
curve that departs from zero around or slightly above 2 ym aero-
dynamic size and reaches a maximum, according to the flow con-
ditions, between 6 and 10 ym.
In Figure 9 the average actual TB deposition for three sub-
jects27 is plotted as a function of particle size for two respi-
ratory patterns.
Q
i
CO
g
0.8
0.6H
0.4 H
0.2-
Volumetric flow rate, cm3/s
Mean residence time, s
Subject 1
Subject 3
Subject 4
___ Lippmann (mean)
Flow rate 370 - 680 cm3/s
Mean residence time 2.1 s
250 750
42
O •
A A
103
10*
105
cm
3/s
Figure 10. Deposition in the ciliated tracheobronchial region during mouth
breathing, in percent of the aerosol entering the trachea. The
straight lines represent the average and the scatter of Lippmann's^
data while the points are the values obtained by Stahlhofen et al.27
(from Stahlhofen et al.27).
360
-------
It has been pointed out that deposition depends strongly
on the health conditions and increases for smokers and again
for bronchitic patients. Chan and Lippmann26 have proposed a
parameter, called the bronchial deposition size, BDS, derived
by expressing tracheobronchial deposition as a function of the
Stokes number. This was found to be 1.20 cm for healthy non-
smokers, 1.02 for smokers, 0.9 for patients under treatment for
obstructive lung disease, and 0.6 for severely disabled patients.
This effect may be important when establishing sampling
guidelines and in evaluating health hazards of airborne particles
since it points out population groups particularly at risk in
some circumstances.
Alveolar Deposition—
The gas exchange region of the airways is characterized
by a very large surface area and therefore by a small average
particle-to-wall distance and a large cumulative cross-section.
Therefore deposition in the aerodynamic size range is practically
due to gravitational settling.
The alveolar deposition therefore increases with increasing
MRT at constant VFR. Because of the behavior of extrathoracic
and tracheobronchial deposition, alveolar deposition also follows
a bell-shaped curve: the relative maximum is around 3 urn and
can be shifted to smaller sizes both for increasing MRT at con-
stant VFR and for increasing VFR at constant MRT. In the first
case the alveolar deposition values increase since the extra-
thoracic and tracheobronchial deposition do not vary appreciably
and gravitational deposition is more effective. Figure 11, from
Heyder et al.,3 shows this effect in detail.
1.0
0.8
c
_o
£ 0.6
O
55 0.4
O
o.
LLJ
O
0.2
MEAN RESIDENCE TIME, 2-8 s ]
VOLUMETRIC FLOW RATE, 250 cm3/4
PARTICLE DENSITY, 0.91 g/cm3
TOTAL DEPOSITION 8
4
2
ALVEOLAR
DEPOSITION
02468
PARTICLE DIAMETER, idtn
10
Figure 11. Effect of particle size and mean residence time on total and
alveolar deposition (from Heyder et al.3).
361
-------
Instead, the increase in VFR causes a higher deposition
by impaction in the higher regions and therefore transmits a
lower fraction of large particles to the alveolar region and
the effect of increased VFR covers the effect of decreased MRT.
The data of Stahlhofen et al.27 for alveolar deposition
are also summarized in Figure 9 as the average of their three
subjects. These are in good agreement with.Lippmann and Albert's25
and Chan and Lippmann"s data for comparable respiration pat-
terns. These data are plotted together for comparison in Figure
12 (from Chan and Lippmann26). The scatter of these are larger
probably because of the larger number of subjects. Consequently
intersubject variability is more considerable and the respiration
is not as strictly controlled. This, though, makes them more
representative of the range of values that one can find in a
population.
INHALABILITY
Very little is known on the efficiency of intake of the
airways for aerosol particles as a function of particle size,
respiratory conditions, wind speed, and orientation of the head.
The only extended work has been done by Ogden and Birkett.31
Experiments were run on a tailor's dummy in a wind tunnel and
with wind speeds of 0.75 and 2.75 m sec"1 and orientations from
0° to 180°.
c
o
£
H-
O
P
co
O
Q.
LU
CC
_l
O
iii
Lw
•~1
<
0.8
0.6
0.4
0.2
0
• i j i | i i i i [
• CHAN AND LIPPMANN 26
- —
- ° STAHLHOFEN, et al 27
— _
• •
"™ * '9 """
N3*
• *
~ •'••£ ~
- **w^ « * \ ~
***** fi -
i * • i ..,?••
0.01 0.05 0.1 0.5 1 5 10 20
DIFFUSIVE DIAMETER, jum AERODYNAMIC DIAMETER, pm
Figure 12. Alveolar deposition data of Chan and Lippmann26 and of
Stahlhofen et al.27
362
-------
With 2.75 m sec"1 wind speed, already at 5 ym diameter the
aerosol is strongly oversampled for nose and mouth breathing
at 84 cm3 sec"1 VFR, is sampled faithfully at 335 cm3 sec"1,
and undersampled for higher VFR. At lower wind speed the portal
of entry is not as important and the deviation of the sampled
concentration from the true concentration is at most of the order
of 25% above 15 - 20 ym.
Orientation between 0° and 90° is very important: with a
5 £ min"1 VFR at 90° and with a wind speed of 0.75 m sec"1, the
nose can sample 5 ym particles with an efficiency almost halved
with respect to 0°. At a higher wind speed and at 45° the nose
can strongly oversample, while the mouth samples always with
a reduced efficiency.
Ogden and Birkett conclude that if the exposure is averaged
for all wind directions, to simulate a worker uniformly exposed,
the differences largely disappear and for winds between 0.75
and 2.75 m sec"1 and minute volumes between 20 and 40 5, for nose
and mouth breathing, the efficiency ranges as functions of the
aerodynamic diameter are shown in Table 1.
TABLE 1. RANGE OF ENTRY EFFICIENCIES FOR A HEAD RANDOMLY
ORIENTED TO A WIND BETWEEN 0.75 AND 2.75 m sec"1
BREATHING WITH MINUTE VOLUME 20 - 40 LITERS3
Aerodynamic
diameter, ym 0 5 10 15 20 25 30
Efficiency
range, % 100 68-83 46-72 39-67 33-60 31-55 30-52
a From Ogden and Birkett31
SENSITIVITY CALCULATIONS
Particle deposition in the airways depends, as shown, on
many parameters. Most of the measurements refer to laboratory
conditions and their bearing on industrial or environmental hy-
giene is not straightforward. It is therefore interesting to
study the sensitivity of total and regional deposition to aerosol
parameters, respiration pattern, inhalation portal, and individual
variability.
363
-------
Total Deposition
The data used for the sensitivity calculation are:
1) total deposition curves as a function o£ particle size,
respiratory frequency, and tidal volume for mouth and nose
breathing obtained by Heyder and coworkers 1 **'3' 27 in the
range 0.3 to 9 ym;
2) intersubject variability between 0.3 and 1.5 ym and electric
charge deposition values at 0.3 and 0.6 ym up to 70 ele-
mentary charges obtained at our laboratory.1*'17
When not otherwise specified the aerosol was assumed log
normally distributed with a geometric standard deviation Og = 2,
which can be considered reasonable in many working environments;
the reference breathing pattern is 1000 cm3 tidal volume and
15 respirations min"1 (abbreviated 1000/15).
The sensitivity to particle size is presented in Table 2
for mouth and nose breathing, where the data is given as a per-
centage variation of deposition with respect to a 1 um, Og = 2
aerosol. The effect of size is very strong, especially for mouth
breathing, and can reach a factor of 3.
TABLE 2. TOTAL DEPOSITION; EFFECT OF MEDIAN SIZE OF PARTICLES
A% Deposition, A% Deposition,
D, ym mouth breathing nose breathing
0.6
1.0
1.5
2
3
4
6
8
10
-28
_
+42
+81
+146
+193
+ 250
+281
+ 299
-33
„_.
+ 31
+ 52
+ 76
+87
+97
+ 100
+ 103
364
-------
In Table 3 the effect of an increase of ag from 2 to 3 is
shown, as a percentage variation of deposition with respect to
the value at a = 2, for mouth and nose breathing. The effect
is especially remarkable again for mouth breathing. The depend-
ence on a is linear, so any a can be evaluated.
In Table 4, an example of the effect of biological vari-
ability is given: it is remarkable at small sizes since the
absolute value is low. At higher sizes deposition approaches
100% for all the subjects; little room is therefore left for
individual variations.
TABLE 3. TOTAL DEPOSITION: EFFECT OF AN INCREASE OF ag
FROM 2 TO 3
D, vim
0.6
1
1.5
2
3
4
A% Deposition,
mouth breathing
+35
+40
+27
+15
-1
-8
A% Deposition,
nose breathing
+39
+9
-5
-9
-9
-8
TABLE 4. TOTAL DEPOSITION:
POSSIBLE INTERSUBJECT
VARIABILITY
D, ym
0.6
1
1.5
2
3
4
A% Deposition
±34
±26
±20
±17
±12
±9
365
-------
The effect of the respiration pattern is shown, as a percent
variation in deposition with respect to 1000/15, in Table 5 for
mouth breathing and Table 6 for nose breathing.
It can be remarked that 500/15 can be taken as reference
for the respiration at rest, 1000/15 for moderate activity and
1000/30 for strong physical activity. The highest deposition
is found in the case of moderate activity. .
Nose breathing is compared to mouth breathing in Table 7;
the data are the percent variation of nose breathing with respect
to mouth breathing.
Table 8 summarizes the sensitivity calculations for the
various parameters in the specified ranges. For the effect of
electric charges, the increase is referred to the 28 elementary
charges on 0.3 pm particles and 70 charges on 0.6 jam. As was
observed above, unipolar charging is rare, but, since deposition
is by image forces, a comparable increase in efficiency could
take place with aerosols freshly produced by mechanical action
and not neutralized.
TABLE 5. TOTAL DEPOSITION: EFFECT OF BREATHING PATTERN
(VIA MOUTH)
A% Deposition3
TV = 500 cm3, TV = 1000 cm3,
D, ym 15 respirations/min 30 respirations/min
0.6
1
1.5
2
3
4
6
8
10
-10%
-15%
-17%
-18%
-18%
-17%
-14%
-11%
-9%
-44%
-33%
-22%
-15%
-7%
-4%
-1%
-0
+1%
With respect to TV = 1000 cm3, 15 respirations/min,
366
-------
TABLE 6. TOTAL DEPOSITION: EFFECT OF BREATHING PATTERN
(VIA NOSE)
D, lam
0.6
1.0
1.5
2
3
4
6
8
10
% Deposition9
TV = 500 cm3,
15 respirations/min
-30%
-30%
-26%
-21%
-15%
-10%
-5%
-3%
-2%
TV = 1000 cm3,
30 respirations/min
+2%
+5%
+5%
+4%
+2%
+2%
+1%
0
0
With respect to TV = 1000 cm3, 15 respirations/min
TABLE 7. TOTAL DEPOSITION:
PERCENT VARIATION OF NOSE
BREATHING WITH RESPECT TO
MOUTH BREATHING
D, ym
0.6
1.0
1.5
2
3
4
6
8
10
A% Deposition
+100
+ 115
+99
+80
+ 53
+38
+ 21
+14
+ 9
367
-------
Only two values for each parameter are given since they
generally act in a different way for sizes greater or smaller
than 1.5 urn.
In the range below 1.5 ym the inhalation route is the most
sensitive variable, while the others have the same effect and
have to be known with comparable accuracy. Above 2 ym the ruling
parameter is particle size.
Aveolar Deposition
The data available at the time the calculation was performed
were essentially the experimental alveolar deposition curve sum-
marized by Lippmann2 for mouth breathing, and, for nose breath-
ing, the curve calculated by introducing the nose deposition.
In the range below 2 pin the curves were completed with Heyder
et al.'s data14 of total deposition, since from 0.3 to 2 pm total
and alveolar deposition coincide.
TABLE 8. TOTAL DEPOSITION: SUMMARIZED SENSITIVITY TO RELEVANT
AEROSOL AND RESPIRATION PARAMETERS
Particle size
range, urn A% Deposition
Size of particlesa
Geometric standard
deviation3
Electric charge
of particles3
Breathing pattern3
Inhalation route,
nose vs mouth
Intersubject
variability
0.6-1.5
2-10
0.6-1.5
2-10
0.3
0.6
0.6-1.5
2-10
0.6-1.5
2-10
0.6-1.5
2-10
300
30
40
10
35
65
30
15
100
80-10
30
10
Mouth breathing
368
-------
In Table 9 the effect of size is shown, for mouth and nose
breathing respectively; the effect of changing ag from 2 to 3
is shown separately for the inhalation portal in Table 10 and
the percent decrease in alveolar deposition when inhaling through
the nose with respect to mouth breathing is shown in Table 11.
As in the case of total deposition, the parameter sensitivity
is summarized in Table 12. The alveolar region is an organ at
risk for "insoluble" particles (or better, slowly transferable
materials); Table 12 therefore can be taken as an indication
of the uncertainties one faces as far as deposition is concerned
when the aerosol is completely unknown. Alveolar deposition
is less sensitive to size, standard deviation, and inhalation
portal than total deposition.
Unfortunately more data is needed, especially for charged
particles, for a complete assessment; nevertheless it can be
stated that the uncertainties of the inhaling subject (respira-
tion pattern, route of entry, biological variability) are com-
parable to the uncertainties of the aerosol characterization.
Some aspects of these will be dealt with in the following
section, but it should be stressed here that, for a realistic
evaluation of inhalation risks, more data is needed also on
oxygen demand, minute volume, frequency, individual variability
of deposition, and route of inhalation.
TABLE 9. ALVEOLAR DEPOSITION: EFFECT OF MEDIAN SIZE
OF PARTICLES
A% Deposition, A% Deposition,
D, iam mouth breathing nose breathing
0.6
1
1.5
2
3
4
6
8
10
-25
-
+26
+40
+45
+32
0
-33
-54
-12
-
+11
+13
+3
-13
-43
-63
-76
369
-------
TABLE 10. ALVEOLAR DEPOSITION: EFFECT OF AN INCREASE OF
qq FROM 2 to 3
D, ym
0.6
1
1.5
2
3
4
A% Deposition,
mouth breathing
+22
+13
-10
-25
-38
-35
A% Deposition,
nose breathing
+19
+2
-16
-27
-32
-22
TABLE 11. ALVEOLAR DEPO-
SITION: PERCENT VARIATION OF
NOSE BREATHING WITH RESPECT
TO MOUTH BREATHING
D, vim
0.6
1
1.5
2
3
4
6
8
10
A% Deposition
-19
-31
-39
-44
-51
-54
-59
-62
-64
370
-------
POSSIBLE SAMPLING METHODS
At our laboratory we have essentially followed two approaches
to aerosol characterization for inhalation risk assessment, be-
sides the consolidated techniques that are generally in use.
One is based on a simplified size spectrometry that can
be used in the field and the other is based on the simulation
of the total and regional deposition in the human respiratory
tract.
The Inertial Spectrometer
Size characterization in the field is now performed by means
of cascade impactors.32 These are simple instruments and so
well known that any description is redundant here. They have
been improved in recent years and have greatly profited from
theoretical studies.33?3" Nevertheless the usual drawbacks have
only partially been overcome; particle bouncing and reentrainment
are still problems. Vapor supersaturation across the last stages,
which may increase particle mass by condensation, has also to
be considered.35 Cascade centripeters36 and generally virtual
impactors have improved the performance as far as stage loading
is concerned, but the wall losses may become a problem.
Sampling cyclones have been greatly improved3 7~'11 with the
aid of semiempirical theories. Sharp cut-offs have been obtained
together with good classification performance. These are par-
ticularly useful in high concentration environments since par-
ticle bouncing and reentrainment can be avoided.
A new instrument has been recently described1*2 that works
as a continuous spectrometer, has no moving parts, and collects
TABLE 12. ALVEOLAR DEPOSITION: SUMMARIZED SENSITIVITY TO SOME
OF THE RELEVANT PARAMETERS
Particle size
range, ym A% Deposition
Size of particles3
Geometr icastandard
deviation
Inhalation route,
nose vs mouth
0
0
0
.6-1.5
2-10
.6-1.5
2-10
.6-1.5
2-10
25
40
20
30
30
50
Mouth breathing
371
-------
tne whole aerosol on just one filter. It is essentially composed
of a 2-mm deep rectangular channel, with a 90° bend, which is
flushed with clean air. A thin aerosol sheath is injected up-
stream of the bend; v.'hile the streamlines follow the curvature,
the particles, because of their inertia, tend to persist in their
initial velocity at the bend. Therefore the particles depart
from the original streamlines by a distance which is a function
of their aerodynamic diameter. The external wall downstream
of the bend is made of a membrane filter through which air is
sucked. The particles are separated according to size while
airborne; the small separation is strongly magnified by the aero-
dynamic projection to the filter surface. The filter will then
collect the whole aerosol: the particles too small to be diverted
will be collected toward the end of the filter, large particles
are precipitated at the beginning, and the intermediate sizes
are continuously separated, The deposit can then be analyzed
to yield the aerodynamic activity diameter, mass diameter, and
elemental size distribution. It could also be turned into a
real time aerosol spectrometer since the ceiling has no other
function than flow containment and can house appropriate detectors
The total flow rate in the present version is 7 £pm and
the aerosol sampling rate can reach 1/10 of that5 while still
preserving good resolution. A high resolution run with 0.5,
1.14, and 2.01 urn latex spheres sampled separately is shown in
Figure ±3.
Figure 13. Deposition pattern of 0.5, 1.14, and 2.01 urn latex particles
sampled separately with the inertial spectrometer. The singlet
and mu/tip/et fines of 1.14 and 2.01 spheres are clearly visible.
372
-------
An example of a calibration curve is shown in Figure 14.
This spectrometer could meet the proposed sampling criteria
of EPA1*3 by working at a reduced flow rate. The deposit could
then be cut into three sections: sizes greater than 15 ym, be-
tween 15 and 2.5 urn, and below 2.5 ym and each section could
be analyzed for the quantity needed for toxicological evaluation.
In fact membrane filters are of controlled composition and are
an ideal means for environmental chemical analysis. The filter
can be easily stored; the procedure could be greatly simplified
since the filter could be evaluated only in the case, for ex-
ample, when a total particulate sampling would exceed a predeter-
mined investigation level.
For general application to size distribution measurements,
the deposit can be scanned and a signal can be obtained that
is a function of the amount of material contained in each size.
Examples of these could be radiometric scanning in the case of
radioactive materials, X-ray fluorescence scanning, or optical
scanning of the diaphanized filter. Otherwise the filter can
be cut, with the aid of the calibration curve, in equal loga-
rithmic intervals and the amount of material contained in each
E
cc
01
UJ
o
o
cc
111
10
9
8
6
5
4-
3
2
1
0
5 10 15 20 25 30 35 40 45 50
DEPOSITION DISTANCE, mm
Figure 14. Calibration curve of the inertial spectrometer.
373
-------
section analyzed with standard microchemical techniques. The
cumulative or differential distribution can be obtained of the
mass or activity and then the activity median or mass median
aerodynamic diameter can be obtained. For each size fraction
the inhalability and deposition can be readily computed even
with a desk-top programmable microcomputer, for the particular
conditions of the individual involved.
Figure 15 shows the distribution of sodium chloride par-
ticles labelled with uranine as analyzed with fluorescence tech-
niques.
If necessary, the aerosol can be humidified in order to
let the hygroscopic particles grow by water condensation. The
pressure drop across the bend is too low to cause any appreciable
modification of the aerosol before deposition.
99.9
99
CO
09
D
O
I-
z
111
O
cc
HI
Q.
90
80
60
40
20
10
5
0.5 1 2 4 6 10
PARTICLE DIAMETER, urn
Figure 15. Cumulative mass vs. size of a uranine-tagged sodium chloride
aerosol sampled with the inertial spectrometer and determined
by fluorometry.
374
-------
Deposition Simulating Sampler
Another approach has been already described1* and has been
recently developed into a field sampler, taking also into ac-
count the most recent data on total and regional deposition.
It is based on the simulation of the total and regional
deposition of particles in the human respiratory tract.
In this way the size characterization of the aerosol is
not needed; the approach is then simpler than size spectrometry
but it is not as flexible in the sense that a particular respira-
tory pattern has to be assumed as a reference.
Bubbles of aerosols freely rising through a water column
experience a particle deposition by inertia, sedimentation, and
diffusion. The combination of these mechanisms produces a deposi-
tion behavior which is quite close to absolute alveolar deposi-
tion in the respiratory tract. Therefore if, before the bubbler,
a stage is added with a deposition efficiency of the tracheo-
bronchial plus extrathoracic airways, what is captured in each
stage is a measure of the effective regional deposition.
In the present set-up the first stage is composed of a dif-
fusion humidifier and a specially designed cyclone. The sum
of the deposition in the two stages gives the total deposition.
This approach has the advantage of inducing condensation growth
of the particles where appropriate. The present set-up "*** has
been checked only for particles larger than 0.5 pm, that is,
only in the aerodynamic range. It has been previously shown,1*
though, that in the diffusion range particles are captured in
the bubbles by a diffusion mechanism; this might be a more cor-
rect approach than pushing aerodynamic separation to sizes where
aerodynamic effects play no role in particle deposition.
There are several parameters adjustable in order to vary
the deposition efficiency: the height of the water column, by
which the residence time of the bubble in water can be varied;
the number and diameter of the bubbling nozzles, by which the
bubble diameters can be to some extent adjusted and the impaction
efficiency during bubble formation can be affected.
For the evaluation of the deposit there are two possibili-
ties: either the material in each stage is recovered and mea-
sured or parallel lines can be assembled, each containing one
stage more than the other and terminating with a membrane filter.
The difference in the filter content would then be attributed
to the capture of the missing stage or stages.
The drawback of this approach is again that it matches one
particular respiratory pattern at a time and it has to be re-
adjusted to match other conditions; this is essentially equi-
valent to selecting a reference condition or threshold that
always causes some rigidity in the system.
375
-------
In Figure 16 the deposition efficiencies obtained in the
sampler are shown by the experimental points as a function of
the particle size.
For comparison the deposition efficiency in humans3 is shown
with the solid lines, for 1500/15 and mouth breathing.
CONCLUSIONS
The deposition efficiency in the human respiratory tract
depends strongly on respiration pattern, portal of inhalation,
minute volume and individual variability besides aerosol proper-
ties, like particle size, shape, electric charge, and solubility.
Therefore, fixing particular particle sizes as a threshold
has a meaning only if referred to particular breathing situations
and for particular aerosols; consequently the need arises for
safety factors that invariably make the evaluation of the inhala-
tion hazard less and less realistic.
100-
90-
S "
2 70H
w 60
o
2i 50
Q
40
30
20
10 -I
0
TOTAL
EXTRATHORACIC +
TRACHEOBRONCHIAL
234567
AERODYNAMIC DIAMETER, jum
Figure 16. Deposition efficiency in the simulating sampler44 (dots), as
compared with the total alveolar deposition in humans for
mouth breathing, 1500/15,3 (solid line).
376
-------
The more rigorous approach would then be a direct evaluation
of particle size distribution and of other aerosol properties.
These could be introduced into the appropriate inhalability and
deposition functions; then the effective deposition in the real-
istic situation could be computed.
This approach may appear like requesting more resources
than the evaluation of just two or three fractions. The amount
of resources needed can be decreased if total dust is sampled
together with the aerosol for size distribution. The aerosol
could then be characterized only if the total dust sample exceeds
a prefixed investigation level or whenever the circumstances
would suggest to do so.
REFERENCES
1. Morrow, P.E., Chairman. ICRP Task Group Report on Lung
Dynamics: Deposition and Retention Models for Internal
Dosimetry of the Human Respiratory Tract. Health Phys.
12:173-208, 1966.
2. Lippmann, M. Regional Deposition of Particles in the Human
Respiratory Tract. In: Handbook of Physiology, Section
9: Reaction to Environmental Agents. D.H.K. Lee, H.L.
Falk, S.D. Murphy, and S.R. Geiger, eds. Bethesda, MD,
American Physiological Society, 1977. pp. 213-232.
3. Heyder, J., J. Gebhart, and W. Stahlhofen. Inhalation of
Aerosols; Particle Deposition and Retention. Presented
at American Chemical Society Annual Meeting, Symposium on
Aerosol Generation and Exposure Facilities, April, 1979.
4. Melandri, C., and V. Prodi. Simulation of the Regional
Deposition of Aerosols in the Respiratory Tract. Am. Ind.
Hyg. Assoc. J. 32:52-57, 1971.
5. Schroter, R.C., and M.F. Sudlow. Flow Patterns in Models
of the Human Airways. Respir. Physiol. 7:431-455, 1969.
6. Ultman, J.S., and H.S. Blatman. Longitudinal Mixing in
Pulmonary Airways. Analysis of Inert Gas Dispersion in
Symmetric Tube Network Models. Respir. Physiol. 30:349-
367, 1977.
7. Taulbee, D.B., and C.P. Yu. A Theory of Aerosol Deposition
in the Human Respiratory Tract. J. Appl. Physiol. 38:77-
85, 1975.
8. Yu, C.P., and D.B. Taulbee. A Theory for Predicting Respira-
tory Tract Deposition of Inhaled Particles in Man. In:
Inhaled Particles IV, W.H. Walton, ed. Pergamon Press,
Oxford, 1977. pp. 35-46.
377
-------
9. Taulbee, D.B., and C.P. Yu. Theory of Particle Deposition
in the Human Lung. Presented at Annual Meeting, Gesell-
schaft fiir Aerosolforschung, Bad Soden, Germany, 1976.
10. Muir, D.C.F., and C.N. Davies. The Deposition of 0.5 ym
Diameter Aerosols in the Lungs of Man. Ann. Occup. Hyg.
10:161-174, 1967.
11. Heyder, J., J. Gebhart, G. Heigwer, C. Roth, and W. Stahl-
hofen. Experimental Studies of the Total Deposition of
Aerosol Particles in the Human Respiratory Tract. J. Aero-
sol Sci. 4:191-208, 1973.
12. Giacomelli-Maltoni, G., C. Melandri, V. Prodi, and G. Tarroni.
Deposition Efficiency of Monodisperse Particles in the Human
Respiratory Tract. Am. Ind. Hyg. Assoc. J. 33:603-610,
1972.
13. Heyder, J., J. Gebhart, C. Roth, W. Stahlhofen, G. Tarroni,
T. De Zaiacomo, M. Formignani, D. Melandri, and V. Prodi.
Intercomparison of Lung Deposition Data for Aerosol Par-
ticles. J. Aerosol Sci. 9:147-155, 1978.
14. Heyder, J., L. Armbruster, J. Gebhart, E. Grein, and W.
Stahlhofen. Total Deposition of Aerosol Particles in the
Human Respiratory Tract for Nose and Mouth Breathing. J.
Aerosol Sci. 6:311-328, 1975.
15. Swift, D.L., F. Shanty, and J.T. O'Neil. Human Respiratory
Deposition Pattern of Fume-Like Particles. Presented at
American Industrial Hygiene Association Conference, May,
1977.
16. Tarroni, G., C. Melandri, V. Prodi, T. De Zaiacomo, M. Formi-
gnani, and P. Bassi. An Indication on the Biological Vari-
ability of Aerosol Total Deposition in Humans. Presented
at American Industrial Hygiene Association Conference, May,
1979.
17. Melandri, C., V. Prodi, G. Tarroni, M. Formignani, T. De
Zaiacomo, F.G. Bompane, and G. Maestri. On the Deposition
of Unipolarly Charged Particles in the Human Respiratory
Tract. In: Inhaled Particles IV, W.H. Walton, ed. Per-
gamon Press, Oxford, 1977. pp. 193-203.
18. Yu, C.P., and K. Chandra. Precipitation of Submicron Charged
Particles in Human Lung Airways. Bull. Math. Biol. 39:471,
1977.
19. Hounam, R.F., A. Black, and M. Walsh. Deposition of Aerosol
Particles in the Nasopharyngeal Region of the Human Respira-
tory Tract. Nature 221:1254-1255, 1969.
378
-------
20. Lippmann, M. , and R.E. Albert. Deposition and Clearance
of Inhaled Particles in the Human Nose. Ann. Otol. Rhinol.
Laryngol. 79:519-528, 1970.
21. Maertens, A., and W. Jacobi. Die in vivo Bestimmung der
Aerosolteilchen Deposition in Atemtrakt bei Mund-bzw-Nasen
Atmung [The In-Vivo Determination of the Deposition of Aero-
sol Particles in the Respiratory Tract by Mouth or Nose
Breathing]. Presented at Annual Meeting, Gesellschaft fur
Aerosolforschung, Bad Soden, Germany, October, 1973.
22. Rudolf, G., and J. Heyder. Deposition of Aerosol Particles
in the Human Nose. Presented at Annual Meeting, Gesellschaft
fur Aerosolforschung, Bad Soden, Germany, October, 1974.
23. Pattle, R.E. The Retention of Gases and Particles in the
Human Nose. In: Inhaled Particles and Vapors. C.N. Davies,
ed. Pergamon Press, Oxford, 1961. pp. 302-309.
24. Heyder, J., and G. Rudolf. Deposition of Aerosol Particles
in the Human Nose. In: Inhaled Particles IV, W.H. Walton,
ed. Pergamon Press, Oxford, 1975. pp. 107-125.
25. Lippman, M., and R.E. Albert. The Effect of Particle Size
on the Regional Deposition of Inhaled Aerosols in the Human
Respiratory Tract. J. Am. Ind. Hyg. Assoc. 30:257-275,
1969.
26. Chan, T.L., and M. Lippmann. Experimental Measurements
and Empirical Modelling of the Regional Deposition of In-
haled Particles in Humans. Submitted to Am. Ind. Hyg.
Assoc. J.
27. Stahlhofen, W., J. Gebhart, and J. Heyder. Experimental
Determination of the Regional Deposition of Aerosol Par-
ticles in the Human Respiratory Tract. Presented at American
Industrial Hygiene Association Conference, May, 1979.
28. Lippmann, M., R.E. Albert, and H.T. Peterson. The Regional
Deposition of Inhaled Aerosols in Man. In: Inhaled Par-
ticles III, W.H. Walton, ed. Unwin, London, 1971, pp.
105-120.
29. Chan, T.L., R.M. Schreck, and M. Lippmann. Effect of Tur-
bulence on Particle Deposition in the Human Trachea and
Bronchial Airways. Presented at Annual Meeting, American
Institute of Chemical Engineers, November, 1978.
30. Chan, T.L., M. Lippmann, V.R. Cohen, and R.B. Schlesinger.
Effect of Electrostatic Charges on Particle Deposition in
a Hollow Cast of the Human Larynx-Tracheobronchial Tree.
J. Aerosol Sci. 9:463-468, 1978.
379
-------
31. Ogden, T.L., and J.L. Birkett. The Human Head as a Dust
Sampler. In: Inhaled Particles IV. W.H. Walton, ed. Perga-
mon Press, Oxford, 1977. pp. 93-105.
32. May, K.R. The Cascade Impactor: An Instrument for Sampling
Coarse Aerosols. J. Sci. Instrum. 22:187-195, 1945.
33. Marple, V.A., and B.Y.H. Liu. Characteristics of Laminar
Jet Impactors. Environ. Sci. Technol. 8:648-654, 1974.
34. Marple, V.A., and K. Willeke. Inertial Impactors-Theory,
Design and Use. In: Fine Particles, Aerosol Generation,
Measurement, Sampling, and Analysis. B.Y.H. Liu, ed.
Academic Press, New York, 1976. pp. 411-446.
35. Tarroni, G., C. Melandri, V. Prodi, M. Formignani, and T.
De Zaiacomo. Taratura di una Centripeta in Cascata per
Campionamenti di Aerosol in Diverse Condizioni Operative.
[Calibration of a Cascade Centripeter for Sampling Aerosols
under Diverse Operating Conditions.] Presented at Annual
Meeting, Italian Health Physics Society, Bologna, October,
1977.
36. Hounam, R.F., and R.J. Sherwood. The Cascade Centripeter:
a Device for Determining the Concentration and Size Distri-
bution of Aerosols. Am. Ind. Hyg. Assoc. J. 26:122-131,
1965.
37. Chan, T., and M. Lippmann. Particle Collection Efficiencies
of Air Sampling Cyclones: An Empirical Theory. Environ.
Sci. Technol. 11:377-382, 1977.
38. Lippmann, M., and T.L. Chan. Cyclone Sampler Performance.
In: Proceedings: Advances in Particle Sampling and Measure-
ment. W.B. Smith, compiler. EPA-600/7-79-065, U.S. Environ-
mental Protection Agency, Research Triangle Park, NC, 1979.
pp. 30-51.
39. Blachman, M.W., and M. Lippmann. Performance Characteristics
of the Multi-Cyclone Aerosol Sampler. Am. Ind. Hyg. Assoc.
J. 35:311-316, 1974.
40. Smith, W.B., K.M. Gushing, G.E. Lacey, and J.D. McCain.
Particulate Sizing Techniques for Control Device Evaluation.
EPA-650/2-74-102a, U.S. Environmental Protection Agency,
Research Triangle Park, NC, 1975. p. 133.
41. Smith, W.B., and R.R. Wilson. Development and Laboratory
Evaluation of a Five-Stage Cyclone System. EPA-600/7-78-
008, U.S. Environmental Protection Agency, Research Triangle
Park, NC, 1978. p. 59.
380
-------
42. Prodi, V., C. Melandri, G. Tarroni, T. De Zaiacomo, M.
Formignani, and D. Hochrainer. An Inertial Spectrometer
for Aerosol Particles. J. Aerosol Sci. 10:411-420, 1979.
43. Miller, F.J., D.E. Gardner, J.A. Graham, R.E. Lee, W.E.
Wilson, and J.D. Bachmann. Size Considerations for Estab-
lishing a Standard for Inhaled Particles. J. Air Pollut.
Control Assoc. 29:610-615, 1979.
44. Melandri, C., V. Prodi, G. Tarroni, M. Formignani, and T.
De Zaiacomo. Aerosol Sampling by Simulation of the Regional
Deposition in the Human Airways. Presented at Annual Meet-
ing, Gesellschaft fur Aerosolforschung, Diisseldorf, October,
1979.
381
-------
PAPER 20
AEROSOL SAMPLING INLETS AND INHALABLE PARTICLES
BENJAMIN Y.H. LIU
DAVID Y.H. PUI
PARTICLE TECHNOLOGY LABORATORY
MECHANICAL ENGINEERING DEPARTMENT
UNIVERSITY OF MINNESOTA
ABSTRACT
The problem of sampling aerosols from the ambient atmosphere
has been considered from a theoretical point of view. Following
a review of the various samplers and inlets used in ambient sampl-
ing, the factors contributing to high sampling efficiency for
large particles are discussed. It is pointed out that the major
mechanisms for particle loss in sampling inlets are impaction
on external surfaces, and impaction, turbulent deposition, and
sedimentation on internal surfaces. Therefore, an efficient
inlet is one for which these losses are minimized.
Based on these theoretical considerations, a new inlet for
sampling inhalable particles (particles with aerodynamic diameter
of 15 um or less) has been designed, constructed, and tested.
The inlet incorporates an inlet configuration allowing for the
efficient entry of large particles into the inlet opening, fol-
lowed by an impactor to remove the coarse, non-inhalable par-
ticles. The inlet has been found to have essentially wind speed
independent characteristics for wind speeds of up to 9 km/hr,
the maximum wind speed used in the tests. The impactor used
has also been found to have sharp cut-off characteristics with
a sharpness of cut parameter, Og, of 1.18. It is believed that
this particular inlet will meet the requirements of a high effi-
ciency inlet for sampling inhalable particles from the ambient
atmosphere.
INTRODUCTION
Sampling of aerosols from the ambient atmosphere is an im-
portant and necessary first step in atmospheric aerosol measure-
ment. To obtain accurate measurement results, a representative
aerosol sample must be drawn through an inlet into the particle
measuring or collecting device. The sampling inlet is an im-
portant part of any aerosol measuring system and, as such, must
382
-------
be designed with care. An ideal inlet should allow all particles
of interest to enter and arrive at the sensing or collecting
zone of the instrument while excluding rain, snow, insects, plant
matter, and other airborne debris. The performance of the inlet
should also be unaffected by wind up to some maximum speed.
Inhalable particulate matter (IPM) has been defined by the
Environmental Protection Agency1 as particles with aerodynamic
diameters of 15 ym or less. New particulate air quality standards
based upon IPM are being developed by the Agency. Unlike the
conventional high-volume sampler which measures the total sus-
pended particulates (TSP) with an uncertain cut-off size for
large particles, samplers for IPM must have well-defined large
particle cut size characteristics. The development of an inlet
of high efficiency for particles of up to 15 ym aerodynamic diam-
eter, therefore, is essential for the development of new IPM
standards.
In this paper, we will first briefly review the various
approaches to ambient aerosol sampling, with particular emphasis
on the inlet designs. A new inlet for the accurate sampling
of IPM will then be described, together with the criteria of
design and the performance data obtained on this new inlet.
AMBIENT AEROSOL SAMPLING AND INLET DESIGNS
A principal requirement of ambient aerosol sampling is that
the instrument must be protected from rain and snow. This can
be accomplished by means of simple weather-proof housings or
more elaborate inlet designs to achieve high inlet efficiency.
The term "inlet efficiency" is used here for the total trans-
mission efficiency of the inlet system, defined as the fraction
of particles transmitted by the inlet from the ambient atmosphere
to the sensing or collecting zone of the instrument. For an
inlet of a specific design, the inlet efficiency is generally
a function of particle size, wind speed, and sampling flow rate.
High inlet efficiency can be achieved easily for small particles,
viz., those with diameters of a few micrometers or less. How-
ever, for large particles, high inlet efficiency can be obtained
only with careful design.
High-Volume Sampler
Figure 1 is a schematic diagram of the widely used high-
volume, or hi-vol, sampler. The hi-vol sampler is the standard
reference method for TSP (total suspended particulate) measure-
ment in the ambient atmosphere, in terms of which the national
ambient particulate air quality standards are defined.2 In the
standard hi-vol, a 20.3 cm x 25.4 cm (8 in. x 10 in.) filter
is placed horizontally within the main sampler housing, which
is about 29 cm x 36 cm (11% in. x 14 in.) in cross-section.
The clearance area between the main housing and the roof at its
closest point is specified to be 580.5 ± 193.5 cm2 (90 ± 30 in.2).
Recent wind tunnel tests of the hi-vol by McFarland et al.3 showed
383
-------
that the D50 cut size (aerodynamic diameter corresponding to
an inlet efficiency of 50%) is >30 ym at a wind speed of 2 km/hr,
decreasing to about 30 ym at 8 km/hr and 17 ym at 24 km/hr.
In addition, the inlet efficiency was found to depend on wind
direction. Table 1 shows a comparison of the inlet efficiency
of the hi-vol with several other ambient aerosol samplers and
inlets.
Isokinetic Sampling
In principle, isokinetic sampling can be used for atmospheric
sampling. Although capable of high accuracy, isokinetic sampling
is seldom used because the system to achieve true isokinetic
conditions in the ambient atmosphere is necessarily complicated:
the inlet must be directed to face the wind and the suction speed
in the inlet must also be adjusted to match the wind speed.
The latter task is extremely difficult to achieve under variable
wind conditions.
Figure 1. Schematic diagram of the high-volume, or hi-volf sampler.
384
-------
TABLE 1. CHARACTERISTICS OF AMBIENT AEROSOL SAMPLERS
AND INLETS
Sampler or inlet
Reference
D50 or D
(Efficiency)
Wind speed,
km/hr
Hi-vol
McFarland,
(1979) 3
et
al.
>30
30
17
ym
ym
ym
2
8
24
Rotating cowl
Rockwell inlet
Sehmel (1973)"
Willeke and McFeters
(1975)6
25 ym
SRI inverted inlet
Beckman dichotomous
sampler inlet
Sierra dichotomous
sampler inlet
Bird, et al. (1973)
McFarland, et al.
(1979) 3
Wedding and John
(1979)
5 5
5
5
12
12
12
15.
13
10.
11
22
15
10
ym (39%)
ym (30%)
ym (20%)
ym (35%)
ym (12%)
ym (35%)
5 ym
ym
5 ym
ym
ym
ym
ym
9.2
46
68
9.2
46
68
2
8
24
0
5
15
42
In the "rotating cowl" sampler (Figure 2) developed by
Sehmel,1* a wind vane was used to direct the inlet of a 56.6 Jl/min
(20 cfm) cascade impactor against the wind. The suction speed
in the inlet was kept fixed at 1.9 km/hr. Thus, the sampler
operated sub-isokinetically most of the time, except near calm
air conditions. The sampler was used to collect particles up
to 63 ym in diameter in a study of the resuspension of soil par-
ticles by wind.
Inverted Inlet
An inverted inlet may consist of a simple open-face filter
operating face down, as shown in Figure 3a, or an open-face
filter attached to a cylindrical tube, as shown in Figure 3b.
Figure 4 shows an inverted inlet designed and tested by Southern
Research Institute.5 The inlet efficiencies were found to be
39%, 30%, and 20%, respectively, for 5 ym diameter particles
at wind speeds of 9.2, 46, and 68 km/hr. For 12 ym particles,
the efficiencies were 35%, 12%, and 35%, respectively, at the
same wind speeds.
385
-------
SYSTEM
SUPPORT
ARM
HIGH VOLUME SAMPLER
CASCADE IMPACTOR
SPINDLE EXTENSION
WIND — •-
|
X
)
CYLINORICAI / / 1
WIND
DIRECTION
SENSITIVE
ROTATING
COWL
SAMPLE INLET / / \ •>
CYLINDRICAL «1S5^ WIND ORIENTATION
COWL
BODY «™""« TA|L (T|N
APPROXIMATE SCALE
25cm
Figure 2. The "rotating cowl" sampler described by Sehmel (1973)
(o)
(b)
Figure 3. The inverted inlet.
386
-------
Circumferential Side-Entrance Inlet
Figure 5 is an omni-directional inlet of the circumferential
side-entrance type. The inlet, designed and built by Rockwell
International Corp. at the suggestion of the author (Liu) and
his colleagues (K.T. Whitby and V.A. Marple), was used in EPA's
CHAMP (Community Health Air Monitoring Program) stations for
ambient air monitoring. The inlet incorporates a circumferential
impactor to remove coarse particles. Tests by Willeke and McFeters6
showed that, at the sampling flow rate of 1133 fc/min (40 cfm),
the cut-point of the circumferential impactor was 25 ym under
calm air conditions. However, no data on the sampling efficiency
as a function of wind speed are available.
Figure 6 shows another inlet of the circumferential, side-
entrance type. The inlet, designed by McFarland et al., was
intended for the virtual dichotomous sampler.8 The same inlet
design is now used in the commercial dichotomous samplers pro-
duced by Beckman Instruments, Inc. (Process Instrument Div.,
2500 Harbour Blvd., Fullerton, CA 92634) and Sierra Instruments,
Inc. (P.O. Box 909, Carmel Valley, CA 93924). Tests on the
Beckman inlet by McFarland et al.3 showed that the D50 size of
the inlet was 15.5 ym at a wind speed of 2 km/hr, decreasing
to 13 and 10.5 ym, respectively, at wind speeds of 8 and 24 km/hr.
Similar tests by Wedding and John (personal communication) on
the Sierra inlet resulted in D50 cut sizes of 11 ym, 22 ym, 15 ym,
and 10 ym, respectively, at wind speeds of 0, 5, 15, and 42 km/hr.
Wind Shield and Baffles
The idea of using a wind shield or baffle to minimize the
effect of wind on the inlet efficiency was suggested by the
author (Liu) and more fully explored by Agarwal in an internal
report at the University of Minnesota Particle Technology Labora-
tory.9
Consider the impaction of particles on a cylindrical object
of radius, R. The impaction is governed by the Stokes number,
Stw= W T/R (1)
where W is the wind speed, and T is the particle relaxation time
given by
T = 2 a2 PpC/9 y (2)
where a is the particle radius, Pp is the particle density, C
is the slip correction, and y is the gas viscosity. For impac-
tion to occur, the Stokes number must be larger than some critical
value. Thus, by choosing R to be sufficiently large, impaction
can be prevented for particles of a certain size up to a certain
maximum wind speed.
387
-------
INTERIOR SURFACES
TEFLON COATED
BOLTS, 120° APART
RAINSHIELD
DECK PLATE
LOCK NUT
SOCKET FOR M8 PLUG-IN
(SAME AS NWL SHIP INLET)
Figure 4. The Southern Research Institute inverted inlet.5
HANDLE
COVER
WATER DRAIN
AFTFR FILTER
VACUUM PUMP
Figure 5. The Rockwell circumferential side-entrance inlet.
388
-------
If particles are to be sampled into a tube of a radius
r, and the tube is placed at a right angle to the wind, particle
impaction at the tube entrance will occur if r is small. How-
ever, if the tube is placed within a cylindrical wind shield
or baffle of a radius R, as shown in Figure 7, and R is made
sufficiently large, particle impaction on the wind shield can
be prevented for particles of the same size at the same wind
speed. If particle impaction does not occur on the wind shield,
the aerosol concentration inside the wind shield must be the
same as the concentration outside. Unbiased sampling can then
take place from the relatively calm air region within the wind
shield. However, no systematic study of the wind shield idea
has been made, and the actual performance of the wind shield
remains unknown.
Figure 8 shows the inlet of a virtual dichotomous sampler
described by Dzubay et al.10 incorporating a wind shield. In
this inlet, an auxiliary blower operating at a relatively high
sampling flow rate of 200 &/min draws in air from around the
wind shield into the annular space, forming a downward-moving
jet. Large particles above 20 ym continue to travel down because
of their inertia, while smaller particles are drawn in through
the circumferential slit opening into the vertical sampling pipe,
where they are sampled isokinetically at a rate of 14 £/min into
the dichotomous sampler. No performance data on this inlet were
given.
-H Mem
INLET
FLOW"
V"
•* V *-
)CC
)C
NTERNAL INLET
•-FLOW
STILLING
"CHAMBER
INLET
"FLOW
(b)
OUTLET
FLOW
Figure 6. The McFarland et al. circumferential side-entrance inlet.7
389
-------
THEORETICAL CONSIDERATIONS
For an aerosol inlet to be efficient, particle loss must
be small. Particle loss can occur on the external surfaces of
an inlet or on internal surfaces. Therefore, an efficient inlet
is one for which both the external and internal particle losses
are minimized.
Consider, for instance, the circumferential, side-entrance
inlet shown schematically in Figure 9. With horizontal wind,
particle impaction can occur on the external surfaces of the
inlet because the streamlines entering the inlet must make a
rather sharp turn at the entrance, and the particle trajectories
can deviate from the streamlines of the flow. This is depicted
in the upper half of the figure. If the inlet is placed at a
distance which is not too far from the instrument housing, stream-
lines deflected by the instrument housing can also enter the
inlet at a rather sharp angle, causing additional impaction
losses to occur. This is depicted in the lower half of the
figure. Therefore, in designing aerosol inlets, both the ex-
ternal shape of the inlet, as well as the distance between the
inlet and the main instrument housing, must be considered.
—WINDSHIELD
SAMPLING
INLET
WIND-
-WINDSHIELD
-SAMPLING
INLET
Figure 7. Wind shield or baffle.
390
-------
Once the particles are brought into an inlet by the stream-
lines, additional loss can occur by impaction, turbulent deposi-
tion, and sedimentation. Careful consideration of all three
internal particle loss mechanisms must be made in order to arrive
at a satisfactory design. Loss of particles by Brownian dif-
fusion is generally unimportant in aerosol inlets because of
the large particle size involved.
DEFLECTOR
RIM
o-* o
LU
INLET
WIND
SHIELD
ANNULAR
VIRTUAL
IMPACTOR
200£/mirT
TO BLOWER
14 C/min
TO PARTICLE SEPARATOR
Figure 8. The wind shield inlet for the virtual dichotomous sampler. 10
391
-------
A completely rigorous theoretical approach to inlet design
does not exist at the present time. Much of the previous theo-
retical work on aerosol inlets was concerned with sampling under
calm air conditions,11"15 and thus was not applicable to ambient
sampling under wind. Similarly, the analysis of Davies16 and
Davies and Subari17 on the sampling efficiency of a thin-walled
tube in a cross-wind, while interesting in showing the mechanics
of sampling, is not particularly useful in inlet designs, since
the straight tube is known to perform poorly under moderate to
high wind conditions and is thus not suitable as a general pur-
pose high efficiency inlet. Thus, instrument designers, when
faced with the problem of inlet design in the past, traditionally
took a rather empirical approach, and many of the resulting
inlets consequently have not performed too well.
In the following discussions, we will show that it is not
necessary to have a completely rigorous theoretical solution
to the problem before an efficient inlet can be designed. We
will show that, with a qualitative understanding of the fluid
and particle mechanics involved, a satisfactory inlet can be
designed.
WIND
— STREAMLINE
— PARTICLE
TRAJECTORY
Figure 9. Streamlines and particle trajectories around an aerosol inlet.
392
-------
Consider the three inlet configurations shown in Figure 10.
A simple vertical tube is used to sample the ambient aerosol
into the instrument. With horizontal wind, particle im-
paction near the upstream side of the tube can occur because
of the bending of the streamlines there. In Figure lOb, a cir-
cular flange is placed over the top of the vertical tube to
deflect the streamlines which would otherwise enter the tube
from below. The flange keeps the streamlines straight prior
to their entry into the tube, thus eliminating particle loss
by external impaction. Therefore/ the design of Figure lOb is
inherently a more efficient design than that of Figure lOa.
Similarly, in Figure lOc, the horizontal entrance near the top
is enlarged compared to that in Figure lOb. This causes the
streamlines to bend more gradually while entering the inlet,
resulting in reduced impaction loss on the internal surfaces.
Thus, of the three inlet designs shown in Figure 10, the design
of Figure lOc should be the most efficient.
The inlet configuration of Figure lOb, and, by extension,
that of Figure lOc, are similar to the case analyzed theoreti-
cally by Zebel.18 Zebel showed that, for an inlet consisting
of a slit or hole in an infinite wall, the aspiration efficiency
can be approximated by the equation
Ei ~
1.09St
(3)
w
WIND
(a)
I
(b)
(0
Figure 10. Comparison of three aerosol inlets.
393
-------
where the aspiration efficiency is the ratio of the number of
particles carried by the flow through the inlet opening to the
number of particles in the original air volume in the ambient
atmosphere, and Stw is the Stokes number based on wind velocity.
In Zebel's calculation, potential flow was assumed. Figure 11
shows the streamlines near the entrance to the slit or hole for
the case where the wind speed, W, is equal to the suction speed
in the slit or hole.
Figure 12 shows the performance of a sampling inlet pre-
dicted by Zebel's equation for a circular inlet of 9.2 cm
(3-5/8 in.) diameter. The result shows that, for particles of
15 ym aerodynamic diameter, the aspiration efficiency is essen-
tially 100% at low wind speeds, decreasing to 90% at a wind speed
of 25 km/hr and 80% at 57 km/hr. Thus, Zebel's results suggest
that it should be possible to design an inlet to sample inhalable
particles efficiently. The inlet performance can also be made
reasonably independent of wind speed up to some rather high wind
speeds.
A NEW IPM INLET
On the basis of the above considerations, we have designed
the inlet shown in Figure 13 for sampling inhalable particles.
The specific inlet has been designed for the virtual dichotomous
sampler, and for a sampling flow rate of 1 m3/hr or 16.7 i/min.
Figure 11. Streamlines over a slit or hole inlet in an infinite wall.
394
-------
#80
G60
040 -
c
£
in
'20-
INLET DIAMETER-9.2cm (3
46 10
WIND SPEED, ktn/hr
2O
40 60
Figure 12. Efficiency of the ideal "hole in an infinite wall" inlet.
CIRCULAR COVER
SPACER (3)
SUPPORTING WIRE (3)
IMPACTION NOZZLE
IMPACTION CUP
\ /
-•V-
Figure 13. The University of Minnesota aerosol inlet for inhalable particles.
395
-------
In addition to the several features described above, the inlet
contains a circular top to keep out rain and snow, and an inter-
nal impactor to remove coarse particles above 15 ym. This inlet
has been evaluated in the wind tunnel at various wind speeds
up to 9 km/hr, the maximum wind speed obtainable in our present
wind tunnel. The experiments performed are described below,
together with the data obtained.
DEFINITION OF TERMS
For purposes of describing the characteristics of the inlet
shown in Figure 13, the following efficiencies are defined:
Ej = aspiration efficiency = n^/m,, (4)
E2 = impactor transmission efficiency = m2/m1 (5)
E = total transmission efficiency of the inlet system,
or the inlet efficiency = m2/m0 = EjE2 (6)
where
mj = mass of particles carried by a given volume of flow
through the exit plane of the impactor nozzle
m2 = mass of particles carried by the same volume of flow
through the exit plane of the inlet system, and
mo = mass of particles in the same volume of air in the
ambient atmosphere.
All particle masses referred to above are for particles of a
certain size.
The aspiration efficiency, Eir defined above, is similar
to the aspiration coefficient defined by Davies11 or the suction
coefficient defined by Zebel18 (Equation 3). The only difference
is that the aspiration efficiency is defined in terms of the
particle mass passing through the exit plane of the impactor
nozzle, whereas the aspiration or suction coefficient is usually
defined in terms of the particle mass brought in by the flow
through the inlet opening, in this case, the top of the funnel.
Thus, the aspiration efficiency takes into account the particle
loss in the entrance portion of the impactor nozzle, whereas
the aspiration or suction coefficient does not. However, since
particle loss in the funnel portion of the inlet has been found
to be small for particles in the inhalable size range, the dif-
ference between aspiration efficiency defined here and the aspira-
tion or suction coefficient defined by Davies and Zebel is not
large and, for all practical purposes, can be ignored.
396
-------
The impactor transmission efficiency, E2, defined by Equa-
tion 5, is usually referred to as "penetration" in impactor
studies. However, in the present case, the impactor is used
to remove the non-inhalable particles and to transmit the in-
halable particles efficiently. The term "impactor transmission
efficiency", therefore, appears appropriate. Finally, the total
transmission efficiency of the inlet system, E, or the inlet
efficiency for short, is the same as the "inlet effectiveness"
defined by McFarland et al.3>7 However, the term "efficiency"
is believed to be appropriate, since in engineering usage, ef-
ficiency generally means, and is usually defined as, the ratio
of the desired output to the input. For example, the efficiency
of a power plant is defined as the ratio of the energy output
of the power plant to the energy input. In the present case,
the inlet system may be considered as a particle transmission
device, the desired output being the aerosol transmitted or
delivered to the sampling or measuring instrument, and the input,
the aerosol drawn in from the ambient atmosphere. Therefore,
the term "inlet transmission efficiency", or the "inlet effi-
ciency" for short, is appropriate and should cause no confusion.
EXPERIMENTAL STUDY AND RESULTS
The preliminary tests of the inlet completed were of two
main types. In one type of test, the inlet was evaluated with
an optical particle counter in the wind tunnel and its perfor-
mance compared with several other inlet configurations. The
purpose of these experiments was to compare the relative per-
formance of the several inlet configurations in order to see
if the chosen configuration indeed had the highest aspiration
efficiency. For these experiments, the absolute value of the
aspiration efficiency was not needed. In the second type of
test, the absolute aspiration efficiency of the inlet was mea-
sured, as well as the absolute impactor transmission efficiency
and the absolute inlet efficiency. The measurement was made
by sampling the particles through the inlet and through an iso-
kinetic probe, and comparing the collected particle mass by
fluorometric analysis.
The experiments were performed in the low-speed wind tunnel
facility of the Particle Technology Laboratory. The wind tunnel
has a 50 cm x 50 cm (20 in. x 20 in.) test section and is capable
of a maximum wind speed of 9 km/hr. The projected area of the
inlet in the direction of flow is about 125 cm2. Thus, about
5% of the test section area was blocked by the inlet under test.
A complete description of the wind tunnel facility has been given
by Whitby et al.*'
Figure 14 shows the test setup for measuring the relative
performance of the inlets by means of an optical particle counter
(OPC). Monodisperse aerosols of oleic acid were generated by
the vibrating orifice monodisperse aerosol generator (Berglund
397
-------
and Liu20) and introduced into the wind tunnel. The particles
were then sampled into an optical particle counter (Royco 245,
Royco Instruments, Inc., 141 Jefferson Drive, Menlo Park, CA
94025) through the different test inlets. By comparing the
counts registered by the OPC when different inlets were used,
the relative performance of the inlets could be determined.
For these experiments, the impactor was removed from the inlet
shown in Figure 13, and the entire flow of 16.7 £/min passing
through the inlet was sampled into the Royco 245 counter. The
geometrical diameter of the particle was calculated from the
liquid flow rate, the solution concentration, and the vibrating
frequency in accordance with the method described by Berglund
and Liu. The geometrical diameter was then converted to aero-
dynamic diameter by multiplying by the factor /Pp, where pp =
0.894 g/cm3 is the density of oleic acid.
The inlets tested included the UM inlet of Figure 13 with
the circular top and impactor cup removed—referred to here as
the reference inlet—, a funnel inlet (the UM inlet without the
top, flange, and impactor cup), and two tube inlets consisting
of straight 5.08 cm (2 in.) and 3.81 cm (1.5 in.) diameter tubes
which were placed perpendicular to the horizontal wind in the
wind tunnel. Under steady state conditions, the counts registered
by the OPC over the same time interval represent the relative
efficiency of the inlets tested.
The results of the above experiments are shown in Figures 15,
16, and 17. In these plots, the aspiration efficiency of the
inlets relative to the reference inlet is shown plotted against
wind speed for particle aerodynamic diameters of 12.8, 14.2,
and 15.6 urn. In all cases, the reference inlet was found to
MONODISPERSE
AEROSOL FROM
VIBRATING
ORIFICE
GENERATOR
FILTER (2)
X
PLATE
H i
X
JL h
r*-i
^— HEPA 1 1
rAIR FLO
I NOZZLE
\1_
TO BLOWER
ROYCO 245
OPTICAL
PARTICLE
COUNTER
V
MANOMETER
Figure 14. Test setup for measuring the relative efficiencies of
different inlet configurations.
398
-------
have the highest aspiration efficiency and to show the least
dependence on wind speed. For example, comparing the reference
inlet with the 3.81 cm diameter tube inlet, the aspiration ef-
ficiency of the latter for 12.8 urn diameter particles is 90%
of that of the reference inlet at a wind speed of 3 km/hr. How-
ever, at a wind speed of 9 km/hr, the aspiration efficiency of
the tube inlet drops to only 12% of that of the reference inlet.
It was also found that the addition of the circular top to the
reference inlet did not change its aspiration efficiency. There-
fore, the data obtained above were, for all practical purposes,
the same as the aspiration efficiency of the various test inlets
relative to the complete UM inlet of Figure 13.
The absolute performance of the complete UM inlet of Fig-
ure 13 was evaluated in the wind tunnel with fluorescein-tagged
DOP (dioctyl phthalate) aerosol and by comparison with an iso-
kinetic probe. The aerosol was generated by dissolving DOP and
fluorescein in isopropyl alcohol and spraying the solution through
1.0
0.9
0.8
(A
<0.6
O
2 0.5
u 0.4
0.3
§
i-
SAMPLING
FLOWRATE' I6.7lpm
OREFERENCE
INLET (UM
INLET WITHOUT \\
TOP AND IMPACTOR) \
(BY DEFINITION)
D 9.2 cm. FUNNEL
A 5.1 cm TUBE
O 3.8 cm TUBE
\
369
WIND SPEED, km/hr
Figure 15. Relative performance of several inlet configurations for
12.8 yuri diameter particles.
399
-------
the vibrating orifice droplet generator. The geometrical diam-
eter of the aerosol was similarly calculated from the droplet
generator operating conditions and converted to aerodynamic diam-
eter by multiplying by /pp, where pp = 0.980 g/cm3 is the density
of DOP. Since the amount of fluorescein used was small (about
2% of the DOP mass in most cases), the slight increase in droplet
density due to fluorescein addition was ignored.
In these tests with fluorescein-tagged DOP aerosol, a Milli-
pore filter was mounted directly at the exit of the UM inlet
to sample the particles passing through the inlet. Following
sampling, the fluorescein was extracted from the filter by im-
mersing the filter in 20 to 80 ml of 0.01 N aqueous solution
of NaOH. The solution and filter were then agitated in an ultra-
sonic bath for 2 to 3 minutes. The fluorescein concentration
in the solution was then measured with a fluorometer (Model 110,
G.K. Turner Associates, Palo Alto, CA). The filter at the exit
of the isokinetic probe was similarly analyzed. To measure the
material deposited on the inside surface of the isokinetic probe,
H
J I.I
E
U!
£l.O
J
£
U
1.
^1
0
*• 0.8
b.
bl
Z
|
< O.6
O
S o.
1 0.
U
U.
lii f\
O.;
z
o
5 Q
K O
a.
*
bl (
K
y
0 0
D,.l42,nn
n SAMPLING
X FLOWRATE • 16.7 Ipm
X
\\
A \ Na
\\
\ V
\\
\\
\\
\\.
v\
0 REFERENCE K
INLET (UM
INLET WITHOUT
TOP AND IMPACTOR)
(BY DEFINITION)
D 9.2 cm FUNNEL
A 5.1 cm TUBE
0 3.8 cm TUBE
369
WIND SPEED, km/hr
Figure 16. Relative performance of several inlet configurations
for 14.2 nm diameter particles.
400
-------
a wet cotton swab was used to wipe the inside surface of the
probe clean. The cotton swab was then immersed in 20 ml of wash
solution and similarly agitated in an ultrasonic bath for 2 to
3 minutes to extract the fluorescein from the cotton swab. Gene-
rally, very little material was found on the inside surface of
the isokinetic probe.
To measure the material collected on the inside surface
of the impaction cup, the outside surface was first wiped clean
with a wet cotton swab and 20 ml of wash solution was added to
the cup and ultrasonically agitated. The solution was then
analyzed. The same cotton swab was then used to wipe the re-
maining surfaces in the inlet to determine the wall losses in
the inlet. The cotton swab was then similarly analyzed.
^*
" I.I
z
bj
5'-°
K
SI
K0.9
fe
Si 0.8
111
Z
o
pO.7
<
«
a
80.6
o
—
<
3 0.5
^
5 o.
J
A.
0.;
z
O
5
S 0.2
CL
V)
> 0
Ul
IT O
0 0
0 '15.6 m
* '
SAMPLING
FLOWRATE'16.7 Ipiti
0
<
\\
\ \v
\ \
\ \
\ ^*V^
V tl
\
\ \
\ \
\ •
\\
\\
\ 1
\\
9 REFERENCE V
INLET
-------
From these measurements, the various efficiencies were cal-
culated as follows:
Impactor transmission efficiency = mf/(m. + m + m,) (7)
-------
100
5 K> 20
AERODYNAMIC DIAMETER, fj.m
Figure 18. Experimental transmission efficiency of the impactor used
in the University of Minnesota inlet.
120
IOO
, so
y eo
u
u.
§40
K
z
3 20
0.5
Dp.
O 8.5
D II.O
£. I3.4
10
WIND SPEED, kffl/hr
Figure 19. Experimental aspiration efficiency of the University
of Minnesota inlet.
403
-------
error of the measured aspiration efficiency (about ±10%), the
aspiration efficiency is essentially 100% for particles of 8.5
and 11.0 ym aerodynamic diameter. However, for particles of
13.4 ym aerodynamic diameter, there appears to be some decrease
in aspiration efficiency at the higher wind speeds. This de-
crease in aspiration efficiency is larger than that predicted
by Zebel's theoretical equation. This suggests that the flow
field at the inlet is more complicated than.that assumed by Zebel
in his analysis.
In Figure 20, the overall transmission efficiency of the
inlet is shown. The change in inlet efficiency with wind speed
is very slight and, for all practical purposes, negligible.
Since the maximum wind speed used in these tests was 9 km/hr,
it would be useful to see what changes in inlet efficiency would
be expected theoretically for higher wind speeds. This can be
done by combining the experimental impactor transmission effi-
ciency curve of Figure 18 with the theoretical aspiration effi-
ciency of Equation 3. The result, shown in Figure 20, suggests
that the inlet should work quite well and not show significant
wind speed dependent characteristics for wind speed of 24 km/hr.
In Figure 20, the characteristics of the present inlet are
compared with those of the inlet used in the Beckman dichotomous
sampler (McFarland et al. 3). The present inlet is seen to be
considerably sharper in its cut-off characteristics.
IOO
80
60
20
WIND SPEED, km/hr
A
O
a
EXPERIMENTAL
EXTRAPOLATED
FROM THEORY
BECKMAN INLET
{McFARLAND el ol.,
1979)
i i i i I
\
V
5 IO
AERODYNAMIC DIAMETER.
50
Figure 20. Experimental inlet efficiency for the University of Minnesota inlet
and its comparison with the Beckman dichotomous sampler inlet.
404
-------
The wall loss data for the present inlet are shown in Table 2,
The maximum wall loss observed is 8.1% for particles of 11.0 ym
aerodynamic diameter, and the average wall loss for all the tests
combined is 1.85%.
DISCUSSIONS, CONCLUSIONS, AND FUTURE WORK
When using the inlet described above on an actual aerosol
sampling instrument, the distance between the inlet and the in-
strument is important and must be considered. If the inlet is
installed too close to the main instrument housing, the situation
depicted in Figure 9 could arise where the streamline deflected
upward by the housing could interfere with the operation of the
inlet, thereby degrading its performance. It is recommended
that the distance between the inlet and the top of the main in-
strument housing be kept equal to or larger than the height of
the instrument housing, as shown in Figure 21(a). If it is not
possible or desirable to have this minimum distance of separation
between the inlet and the instrument housing, a horizontal de-
flection plate can be installed over the top of the instrument
housing, as shown in Figure 21(b), to prevent the streamlines
deflected upward by the instrument housing from interfering with
the inlet operation. With these precautions, it is believed
that the performance of the inlet, as determined for an isolated
inlet in a wind tunnel, should be the same as the actual perfor-
mance of the inlet when installed on the instrument.
TABLE 2. WALL LOSS IN UM INLET
Wind speed, km/hr
Da,
Wall loss, %
1 5
8.5
11
13.4
16
18.5
0.9
1.0
8.1
0
0.1
0.9
11
13.4
3.0
0.4
11
13.4
3.8
1.3
405
-------
Based on the work performed so far, we believe the inlet
described above will meet the basic requirement of an inlet for
sampling inhalable particles from the ambient atmosphere. How-
ever, the D50 diameter of the present inlet is 13.3 pm, which
is somewhat smaller than the 15.0 ym diameter that was intended.
The D5o diameter can be increased by increasing the impactor
nozzle diameter slightly, from the present value of 1.28 cm to
1.39 cm. This slight increase in nozzle diameter should not
affect the inlet operation in other respects.
Before the present inlet is adopted for routine sampling,
it is desirable that additional tests be performed. In addi-
tion to wind tunnel testing at higher wind speeds, the inlet
should also be tested for its actual field performance charac-
teristics. Some limited tests have shown that the inlet is
weather-proof, at least for rain, i.e., the circular top is ef-
fective in keeping out rain. We believe that insects, plant
matter and other airborne debris would, by virtue of their large
size, be effectively trapped by the impact ion cup. By coating
the inside surface of the impaction cup with a non-volatile
grease, particle bounce and re-entrainment could also be pre-
vented. Unlike other impactors designed for size classification
of small aerosol particles, the present impactor operates at
a relatively low jet velocity of 8.4 km/hr, which should minimize
the problem of particle bounce and re-entrainment. In addition,
if coarse particles should bounce when they come in contact with
the bottom of the cup, they would lose much of their momentum
-INLET
-INLET
INSTRUMENT
HOUSING
/-DEFLECTOR
-2H-
INSTRUMENT
HOUSING
Figure 21. Two methods of installing an inlet on an
aerosol sampling instrument.
406
-------
and energy and be subsequently caught by the vertical surfaces
in the impaction cup. Therefore, we do not believe that particle
bounce and re-entrainment would be a problem for an impactor
of this particular design.
In addition to the above, we have also designed an inlet
which is similar to the inlet described above, except that a
cyclone is used for separating the inhalable and non-inhalable
particles. This new inlet will be evaluated and reported some-
time in the future.
ACKNOWLEDGEMENTS
This research is supported by a grant, No. R804600, from
the Environmental Protection Agency. The Agency's support is
gratefully acknowledged. We also wish to thank Sandra Iverson
of the University of Minnesota for her able assistance in carry-
ing out some of the experimental measurements reported here.
This report is Particle Technology Laboratory Publication No. 397.
REFERENCES
1. Miller, F.J., D.E. Gardner, J.A. Graham, R.E. Lee, Jr.,
W.E. Wilson, and J.D. Bachmann. Size Considerations for
Establishing a Standard for Inhalable Particles. J. Air
Pollut. Control Assoc. 29:610, 1979.
2. National Primary and Secondary Ambient Air Quality Standards.
Federal Register 36(84} Friday, April 30, Part II, 1971.
3. McFarland, A.R., C.A. Ortiz, and C.E. Rodes. Characteris-
tics of Aerosol Samplers Used in Ambient Air Monitoring.
Presented at the 86th National Meeting of the American Insti-
tute of Chemical Engineers, 1979.
4. Sehmel, G.A. An Evaluation of High-Volume Cascade Particle
Impactor Systems. Presented at the 2nd Joint Conference
on Sensing of Environmental Pollutants, Instrument Society
of America, 1973.
5. Bird, A.N., Jr., D.V. Brady, and J.D. McCain. Evaluation
of Sampling Systems for Use of the M8 Alarm Aboard Ships.
Report SRI-EAS-73-064, Southern Research Institute, Birming-
ham, AL, 1973.
6. Willeke, K., and J.J. McFeters. Calibration of the CHAMP
Fractionator. Particle Technology Laboratory Publication
No. 252, University of Minnesota, 1975.
7. McFarland, A.R., J.B. Wedding, and J.E. Cermak. Wind Tunnel
Evaluation of a Modified Andersen Impactor and an Ail-Weather
Sampler Inlet. Atmos. Environ. 11:535, 1977.
407
-------
8. Dzubay, T.G., and R.K. Stevens. Ambient Air Analysis with
Dichotomous Sampler and X-Ray Fluorescence Spectrometer.
Environ. Sci. Technol. 9:663, 1975.
9. Agarwal, J.K. The Sampling of Aerosols. Particle Technology
Laboratory Report No. 208, University of Minnesota, 1972.
10. Dzubay, T.G., R.K. Stevens, and C.M. Peterson. Application
of the Dichotomous Sampler to the Characterization of Ambient
Aerosols. In: X-Ray Fluorescence Analysis of Environmental
Samples, T.G. Dzubay, ed. Ann Arbor Science Publishers,
Ann Arbor, MI, 1977. p. 95.
11. Davies, C.N. The Entry of Aerosols into Sampling Tubes
and Heads. Brit. J. Appl. Phys. Ser. 2, 1:921, 1968.
12. Kaslow, D.E., and R.J. Emrich. Aspirating Flow Pattern
and Particle Inertia Effects Near a Blunt Thick-Walled Tube
Entrance. Department of Physics, Lehigh University, Beth-
lehem, PA. Technical Report No. 23, 1973.
13. Kaslow, D.E., and R.J. Emrich. Particle Sampling Efficiencies
for an Aspirating Blunt Thick-Walled Tube in Calm Air.
Department of Physics, Lehigh University, Bethlehem, PA.
Technical Report No. 25, 1974.
14. Kim, Y.W. An Analytical Consideration of the Particle In-
ertia Effect with an Application to Aerosol Sampling Ef-
ficiency Calculation. Department of Physics, Lehigh Univer-
sity, Bethlehem, PA. Technical Report No. 24, 1974.
15. Agarwal, J.K., and B.Y.H. Liu. A Criterion for Accurate
Aerosol Sampling in Calm Air. To be published in Am. Ind.
Hyg. Assoc. J.
16. Davies, C.N. Sampling Aerosols with a Thin-Walled Tube.
Presented at the 12th International Colloquium on Polluted
Atmospheres, Paris, May, 1976.
17. Davies, C.N., and M. Subari. Inertia Effects in Sampling
Aerosols. In: Proceedings: Advances in Particle Sampl-
ing and Measurement. W.B. Smith, compiler. EPA-600-7-79-
065, U.S. Environmental Protection Agency, Research Triangle
Park, NC, 1979.
18. Zebel, G. Some Problems in the Sampling of Aerosols. In:
Recent Developments in Aerosol Science, D.T. Shaw, ed.
Wiley, New York, 1978. p. 167.
19. Whitby, K.T., A.B. Algren, R.C. Jordan, and J.C. Annis.
Evaluation of Air Cleaners for Air Conditioning and Ventila-
tion, Part I—Apparatus. ASHAE Trans. 64:401, 1958.
20. Berglund, R.N., and B.Y.H. Liu. Generation of Monodisperse
Aerosol Standards. Environ. Sci. Technol. 7:141, 1973.
408
-------
METRIC CONVERSION FACTORS
To convert from;
pounds, avoirdupois (Ib)
grains (gr)
grains/cubic foot
(gr/ft3)
inches (in.)
feet (ft)
feet/minute (ft/min)
cubic feet/minute
(ft3/min or cfm)
gallons (U.S.) (gal.)
gallons (U.S.)/minute
(gal./min)
gallons/1000 cubic feet
(gal./lOOO ft3)
inches water gauge
(in. WG or H2O)
temperature °F
To;
kilograms (kg)
grams (g)
grams/cubic meter
(g/m3)
centimeters (cm)
meters (m)
meters/second (m/s)
cubic meters/second
(m3/s)
liters
liters/second (i/s)
liters/cubic meter
/ o /«. 3 \
millimeters of mercury
(mm Hg)
temperature °C
Multiply by;
0.454
0.0648
2.29
2.54
3.05
0.508
0.000472
3.79
0.0632
0.134
1.87
(°F-32) x 5/9
409
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/9-80-004
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Proceedings: Advances in Particle Sampling and
Measurement (Daytona Beach, FL, October 1979)
5. REPORT DATE
January 1980
6, PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
W.B. Smith, Editor
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
10. PROGRAM ELEMENT NO.
EHE624
11. CONTRACT/GRANT NO.
68-02-3118
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Proceedings; 4-11/79
14. SPONSORING AGENCY CODE
EPA/600/13
15.SUPPLEMENTARYNOTESIERL_RTP project officer is D. Bruce Harris, Mail Drop 62,
919/541-2557.
is. ABSTRACT Tne proceedings consist of 20 reports of research on equipment and tech-
niques for sampling and characterizing particulate emissions and other aerosols.
The inhalable particle size range (up to 15 micrometers) is emphasized, and the
basis for selecting this range as a standard is discussed. Novel or improved equip-
ment includes: virtual impactors; impactors for sampling high dust loadings; an
impactor/quartz-crystal-microbalance combination used to sample stratospheric
aerosols; a tapered-element oscillating microbalance for monitoring particulate
emissions and aerosols; an automated piezoelectric microbalance for monitoring
atmospheric aerosols; a hot-wire probe for measuring liquid droplets; sampling
systems that are improvements on EPA Method 5 equipment for measuring mass
emissions; and more efficient sampling probe inlets. New or improved techniques
include: measurement of aerodynamic diameter by laser/doppler velocimetry of
particles accelerated in a converging nozzle; automation of diffusion-battery/conden-
sation nucleus counter systems; sampling inhalable particles in fugitive aerosols;
particle-size spectrometry for characterizing inhalation toxicity; computer extrapo-
lation of particle-size ranges; and the identification of impactor errors due to non-
ideal behavior to particle deposition in sampling probe nozzles.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Pollution
Dust
Aerosols
Measurement
Sampling
Properties
Analyzing
Pollution Control
Stationary Sources
Particulate
Characterizing
13B
11G
07D
14B
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
418
2O. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
410
------- |